mirror of https://github.com/YifanXu74/MQ-Det.git
init
YifanXu74
commit
6fc1e46e79
.asset
configs
odinw_13
odinw_35
groundingdino_new
datasets
models
GroundingDINO
Binary file not shown.
|
After Width: | Height: | Size: 974 KiB |
|
|
@ -0,0 +1,13 @@
|
|||
build/
|
||||
__pycache__/
|
||||
DATASET/
|
||||
MODEL/
|
||||
OUTPUT/
|
||||
.vscode/
|
||||
*.egg-info/
|
||||
*gnu.so
|
||||
logs/
|
||||
odinw-img-logs/
|
||||
other_log/
|
||||
vs_downstream_log/
|
||||
vs_final_downstream_log/
|
||||
|
|
@ -0,0 +1,74 @@
|
|||
We provide guidance for preparing the data used by MQ-DET. Note that not all data are needed for a specific experiments. Please check the `` Required Data`` fields in [README](README.md) to download necessary data. All data should by placed under the ``DATASET`` folder.
|
||||
|
||||
The data should be organized in the following format:
|
||||
```
|
||||
DATASET/
|
||||
coco/
|
||||
annotations/
|
||||
lvis_od_train.json
|
||||
lvis_od_val.json
|
||||
lvis_v1_minival_inserted_image_name.json
|
||||
train2017/
|
||||
val2017/
|
||||
test2017/
|
||||
Objects365/
|
||||
images/
|
||||
zhiyuan_objv2_train.json
|
||||
odinw/
|
||||
AerialMaritimeDrone/
|
||||
...
|
||||
WildfireSmoke/
|
||||
```
|
||||
|
||||
|
||||
|
||||
#### ``Objects365``
|
||||
We found that the Objects365 v1 is unavailable now. Please try to download v2 as follows.
|
||||
|
||||
Download the [Objects365](https://www.objects365.org/overview.html) dataset from [YOLOv5](https://github.com/ultralytics/yolov5/blob/master/data/Objects365.yaml).
|
||||
|
||||
You can also use custom datasets for modulated pre-training as long as they are in COCO format.
|
||||
|
||||
|
||||
#### ``LVIS``
|
||||
LVIS use the same images as COCO. Thus prepare the COCO images and annoations first and place them at ``DATASET/coco/``.
|
||||
|
||||
**All processed LVIS annotation files can be downloaded through:**
|
||||
|
||||
|train|minival|val 1.0|
|
||||
|-----|-------|-------|
|
||||
|[link](https://drive.google.com/file/d/1UpLRWfvXnGrRrhniKuiX_E1bkT90yZVE/view?usp=sharing)|[link](https://drive.google.com/file/d/1lLN9wole5yAsatFpYLnlnFEgcbDLXTfH/view?usp=sharing)|[link](https://drive.google.com/file/d/1BxlNOXEkcwsY2w2QuKdA2bdrrKCGv08J/view?usp=sharing)|
|
||||
|
||||
And place them at ``DATASET/coco/annotations/``.
|
||||
|
||||
|
||||
**If you want to process by yourself rather than using the pre-processed files**, please follow the [instruction in GLIP](https://github.com/microsoft/GLIP/blob/main/DATA.md), summarized as following.
|
||||
|
||||
Download the following annotation files:
|
||||
```
|
||||
wget https://penzhanwu2bbs.blob.core.windows.net/data/GLIPv1_Open/coco/annotations/lvis_v1_minival_inserted_image_name.json -O DATASET/coco/annotations/lvis_v1_minival_inserted_image_name.json
|
||||
|
||||
wget https://penzhanwu2bbs.blob.core.windows.net/data/GLIPv1_Open/coco/annotations/lvis_od_val.json -O coco/annotations/lvis_od_val.json"
|
||||
```
|
||||
Also download the training set for extracting vision queries:
|
||||
```
|
||||
wget https://s3-us-west-2.amazonaws.com/dl.fbaipublicfiles.com/LVIS/lvis_v1_train.json.zip -O coco/annotations/lvis_v1_train.json.zip
|
||||
```
|
||||
Unpack the .zip file to ``coco/annotations/lvis_v1_train.json``, and convert it to coco format:
|
||||
```
|
||||
|
||||
python utils/add_file_name.py
|
||||
```
|
||||
|
||||
|
||||
|
||||
#### ``Object Detection in the Wild (ODinW)``
|
||||
|
||||
**Download ODinW**
|
||||
```
|
||||
python odinw/download_datasets.py
|
||||
```
|
||||
|
||||
``configs/odinw_35`` contain all the meta information of the datasets. ``configs/odinw_13`` are the datasets used by GLIP. Each dataset follows the coco detection format.
|
||||
|
||||
Please refer to [GLIP](https://github.com/microsoft/GLIP/tree/main) for more details.
|
||||
|
|
@ -0,0 +1,54 @@
|
|||
**fatal error: THC/THC.h: No such file or directory**
|
||||
It is because pytorch removed THC/THC.h after its version 1.11. One solution is to downgrade the torch version, but this may be incompatible with the system dependencies (e.g., GPUs, CUDA, ...)
|
||||
Another solution is to modify the cuda file:
|
||||
1. remove all #include <THC/THC.h>
|
||||
2. replace all
|
||||
```
|
||||
THCudaCheck(...);
|
||||
```
|
||||
with
|
||||
```
|
||||
AT_CUDA_CHECK(...);
|
||||
```
|
||||
**THCCeilDiv is undefined**
|
||||
1. #include <ATen/ceil_div.h>
|
||||
2. replace all
|
||||
```
|
||||
THCCeilDiv(...)
|
||||
```
|
||||
with
|
||||
```
|
||||
at::ceil_div(...)
|
||||
```
|
||||
**THCudaMalloc/THCudaFree/THCState is undefined**
|
||||
1. #include <ATen/cuda/ThrustAllocator.h>
|
||||
2. remove the line with THCState
|
||||
3. replace
|
||||
```
|
||||
THCudaMalloc(param1, param2)
|
||||
```
|
||||
with
|
||||
```
|
||||
c10::cuda::CUDACachingAllocator::raw_alloc(param2)
|
||||
```
|
||||
4. replace
|
||||
```
|
||||
THCudaFree(param1, param2)
|
||||
```
|
||||
with
|
||||
```
|
||||
c10::cuda::CUDACachingAllocator::raw_delete(param2)
|
||||
```
|
||||
|
||||
**unrecognized arguments: --local-rank=5**
|
||||
This is because torch with a high version receive ``--local-rank`` rather than ``--local_rank``.
|
||||
Replace ``--local-rank`` with ``--local_rank`` in coresponding code, and vice versa.
|
||||
|
||||
**ImportError: libGL.so.1: cannot open shared object file: No such file or directory**
|
||||
Solved by this [link](https://stackoverflow.com/questions/55313610/importerror-libgl-so-1-cannot-open-shared-object-file-no-such-file-or-directo)
|
||||
|
||||
**Error in dataloader**
|
||||
Try to pass:
|
||||
```
|
||||
DATALOADER.NUM_WORKERS 0
|
||||
```
|
||||
|
|
@ -0,0 +1,252 @@
|
|||
# MQ-Det: Multi-modal Queried Object Detection in the Wild (NeurIPS2023)
|
||||
<!-- <img src=".asset/method.png" width="800"> -->
|
||||
|
||||
Official PyTorch implementation of "[MQ-Det: Multi-modal Queried Object Detection in the Wild](https://arxiv.org/abs/2305.18980)": the first multi-modal queried open-set object detector.
|
||||
|
||||
## Citation
|
||||
|
||||
If you find our work useful in your research, please consider citing:
|
||||
```
|
||||
@article{mqdet,
|
||||
title={Multi-modal queried object detection in the wild},
|
||||
author={Xu, Yifan and Zhang, Mengdan and Fu, Chaoyou and Chen, Peixian and Yang, Xiaoshan and Li, Ke and Xu, Changsheng},
|
||||
journal={Advances in Neural Information Processing Systems},
|
||||
year={2023}
|
||||
}
|
||||
```
|
||||
|
||||
|
||||
## Multi-modal Queried Object Detection
|
||||
We introduce **MQ-Det**, an efficient architecture and pre-training strategy design to utilize both textual description with open-set generalization and visual exemplars with rich description granularity as category queries, namely, **M**ulti-modal **Q**ueried object **Det**ection, for real-world detection with both open-vocabulary categories and various granularity.
|
||||
|
||||
<img src=".asset/method.png" width="800">
|
||||
|
||||
## Method
|
||||
MQ-Det incorporates vision queries into existing well-established language-queried-only detectors.
|
||||
|
||||
**Features**:
|
||||
|
||||
- A plug-and-play gated class-scalable perceiver module upon the frozen detector.
|
||||
- A vision conditioned masked language prediction strategy.
|
||||
- Compatible with most language-queried object detectors.
|
||||
|
||||
<!-- ## TODO
|
||||
|
||||
- [x] Release finetuning-free inference code.
|
||||
- [x] Release checkpoints.
|
||||
- [x] Release fine-tuning code.
|
||||
- [x] Release modulated training code.
|
||||
- [ ] More detailed instruction on applying MQ-Det to custom language-queried detectors. -->
|
||||
|
||||
|
||||
|
||||
|
||||
## Preparation
|
||||
|
||||
**Environment.**
|
||||
Init the environment:
|
||||
```
|
||||
git clone https://github.com/YifanXu74/MQ-Det.git
|
||||
cd MQ-Det
|
||||
conda create -n mqdet python=3.9 -y
|
||||
conda activate mqdet
|
||||
bash init.sh
|
||||
```
|
||||
|
||||
The implementation environment in the paper is python==3.9, torch==2.0.1, GCC==8.3.1, CUDA==11.7. Several potential errors and their solutions are presented in [DEBUG.md](DEBUG.md)
|
||||
<!-- THC/THC.h error with high torch version can be solved by [link1](https://github.com/NVIDIA/DeepLearningExamples/issues/1090) and [link2](https://aitechtogether.com/python/76425.html) -->
|
||||
|
||||
**Data.** Prepare ``Objects365`` (for modulated pre-training), ``LVIS`` (for evaluation), and ``ODinW`` (for evaluation) benchmarks following [DATA.md](DATA.md).
|
||||
|
||||
**Initial Weight.** MQ-Det is build upon frozen language-queried detectors. To conduct modulated pre-training, download corresponding pre-trained model weights first.
|
||||
|
||||
We apply MQ-Det on [GLIP](https://github.com/microsoft/GLIP) and [GroundingDINO](https://github.com/IDEA-Research/GroundingDINO):
|
||||
|
||||
```
|
||||
GLIP-T:
|
||||
wget https://huggingface.co/GLIPModel/GLIP/resolve/main/glip_tiny_model_o365_goldg_cc_sbu.pth -O MODEL/glip_tiny_model_o365_goldg_cc_sbu.pth
|
||||
GLIP-L:
|
||||
wget https://huggingface.co/GLIPModel/GLIP/resolve/main/glip_large_model.pth -O MODEL/glip_large_model.pth
|
||||
GroundingDINO-T:
|
||||
wget https://huggingface.co/ShilongLiu/GroundingDINO/resolve/main/groundingdino_swint_ogc.pth -O MODEL/groundingdino_swint_ogc.pth
|
||||
```
|
||||
If the links fail, please manually download corresponding weights from the following table or the github pages of [GLIP](https://github.com/microsoft/GLIP)/[GroundingDINO](https://github.com/IDEA-Research/GroundingDINO).
|
||||
|
||||
|GLIP-T|GLIP-L|GroundingDINO-T|
|
||||
|------|------|------|
|
||||
|[weight](https://huggingface.co/GLIPModel/GLIP/resolve/main/glip_tiny_model_o365_goldg_cc_sbu.pth)|[weight](https://huggingface.co/GLIPModel/GLIP/resolve/main/glip_large_model.pth)|[weight](https://huggingface.co/ShilongLiu/GroundingDINO/resolve/main/groundingdino_swint_ogc.pth)|
|
||||
|
||||
The weight files should be placed as follows:
|
||||
|
||||
```
|
||||
MODEL/
|
||||
glip_tiny_model_o365_goldg_cc_sbu.pth
|
||||
glip_large_model.pth
|
||||
groundingdino_swint_ogc.pth
|
||||
```
|
||||
|
||||
## Model Zoo
|
||||
The table reports the finetuning-free performance with 5 vision queries.
|
||||
Model | LVIS MiniVal | LVIS Val v1.0 | ODinW-13 | ODinW-35 | Config | Weight
|
||||
-- | -- | -- | -- | -- | -- | --
|
||||
MQ-GLIP-T | 30.4 | 22.6 | 45.6 | 20.8 | [config](configs/pretrain/mq-glip-t.yaml) | [weight](https://drive.google.com/file/d/1n0_D-tisqN5v-IESUEIGzMuO-9wolXiu/view?usp=sharing)
|
||||
MQ-GLIP-L | 43.4 | 34.7 | 54.1 | 23.9 | [config](configs/pretrain/mq-glip-l.yaml) | [weight](https://drive.google.com/file/d/1O_eb1LrlNqpEsoxD23PAIxW8WB6sGoBO/view?usp=sharing)
|
||||
|
||||
## Vision Query Extraction
|
||||
|
||||
**Take MQ-GLIP-T as an example.**
|
||||
|
||||
If you wish to extract vision queries from custom dataset, specify the ``DATASETS.TRAIN`` in the config file.
|
||||
We provide some examples in our implementation in the following.
|
||||
|
||||
### Objects365 for modulated pre-training:
|
||||
|
||||
```
|
||||
python tools/extract_vision_query.py --config_file configs/pretrain/mq-glip-t.yaml --dataset objects365 --add_name tiny
|
||||
```
|
||||
This will generate a query bank file in ``MODEL/object365_query_5000_sel_tiny.pth``
|
||||
|
||||
|
||||
### LVIS for downstream tasks:
|
||||
```
|
||||
python tools/extract_vision_query.py --config_file configs/pretrain/mq-glip-t.yaml --dataset lvis --num_vision_queries 5 --add_name tiny
|
||||
```
|
||||
This will generate a query bank file in ``MODEL/lvis_query_5_pool7_sel_tiny.pth``.
|
||||
|
||||
|
||||
### ODinW for downstream tasks:
|
||||
|
||||
```
|
||||
python tools/extract_vision_query.py --config_file configs/pretrain/mq-glip-t.yaml --dataset odinw-13 --num_vision_queries 5 --add_name tiny
|
||||
```
|
||||
This will generate query bank files for each dataset in ODinW in ``MODEL/{dataset}_query_5_pool7_sel_tiny.pth``.
|
||||
|
||||
### Some paramters corresponding to the query extraction:
|
||||
|
||||
``DATASETS.FEW_SHOT``: if set ``k>0``, the dataset will be subsampled to k-shot for each category when initializing the dataset. This is completed before training. Not used during pre-training.
|
||||
|
||||
``VISION_QUERY.MAX_QUERY_NUMBER``: the max number of vision queries for each category when extracting the query bank. Note that the query extraction is conducted before training and evaluation.
|
||||
|
||||
``VISION_QUERY.NUM_QUERY_PER_CLASS`` controls how many queries to provide for each category during one forward process in training and evaluation.
|
||||
|
||||
Usually, we set
|
||||
|
||||
``VISION_QUERY.MAX_QUERY_NUMBER=5000``, ``VISION_QUERY.NUM_QUERY_PER_CLASS=5``, ``DATASETS.FEW_SHOT=0`` during pre-training.
|
||||
|
||||
``VISION_QUERY.MAX_QUERY_NUMBER=5``, ``VISION_QUERY.NUM_QUERY_PER_CLASS=5``, ``DATASETS.FEW_SHOT=5`` during few-shot (5-shot) fine-tuning.
|
||||
|
||||
``--num_vision_queries`` denotes number of vision queries for each category, and can be an arbitrary number. This will set both ``VISION_QUERY.MAX_QUERY_NUMBER`` and ``DATASETS.FEW_SHOT`` to ``num_vision_queries``.
|
||||
Note that here ``DATASETS.FEW_SHOT`` is only for accelerating the extraction process.
|
||||
|
||||
``--add_name`` is only a mark for different models.
|
||||
For training/evaluating with MQ-GLIP-T/MQ-GLIP-L/MQ-GroundingDINO, we set ``--add_name`` to 'tiny'/'large'/'gd'.
|
||||
|
||||
## Modulated Training
|
||||
|
||||
**Take MQ-GLIP-T as an example.**
|
||||
|
||||
```
|
||||
python -m torch.distributed.launch --nproc_per_node=8 tools/train_net.py --config-file configs/pretrain/mq-glip-t.yaml --use-tensorboard OUTPUT_DIR 'OUTPUT/MQ-GLIP-TINY/'
|
||||
```
|
||||
To conduct pre-training, one should first extract vision queries before start training following the above [instruction](#vision-query-extraction).
|
||||
To pre-train on custom datasets, please specify ``DATASETS.TRAIN`` and ``VISION_SUPPORT.SUPPORT_BANK_PATH`` in the config file.
|
||||
|
||||
## Finetuning-free Evaluation
|
||||
**Take MQ-GLIP-T as an example.**
|
||||
|
||||
### LVIS Evaluation
|
||||
MiniVal:
|
||||
```
|
||||
python -m torch.distributed.launch --nproc_per_node=4 \
|
||||
tools/test_grounding_net.py \
|
||||
--config-file configs/pretrain/mq-glip-t.yaml \
|
||||
--additional_model_config configs/vision_query_5shot/lvis_minival.yaml \
|
||||
VISION_QUERY.QUERY_BANK_PATH MODEL/lvis_query_5_pool7_sel_tiny.pth \
|
||||
MODEL.WEIGHT ${model_weight_path} \
|
||||
TEST.IMS_PER_BATCH 4
|
||||
```
|
||||
Val 1.0:
|
||||
```
|
||||
python -m torch.distributed.launch --nproc_per_node=4 \
|
||||
tools/test_grounding_net.py \
|
||||
--config-file configs/pretrain/mq-glip-t.yaml \
|
||||
--additional_model_config configs/vision_query_5shot/lvis_val.yaml \
|
||||
VISION_QUERY.QUERY_BANK_PATH MODEL/lvis_query_5_pool7_sel_tiny.pth \
|
||||
MODEL.WEIGHT ${model_weight_path} \
|
||||
TEST.IMS_PER_BATCH 4
|
||||
```
|
||||
Please follow the above [instruction](#vision-query-extraction) to extract corresponding vision queries. Note that `--nproc_per_node` must equal to `TEST.IMS_PER_BATCH`.
|
||||
|
||||
### ODinW / Custom Dataset Evaluation
|
||||
```
|
||||
python tools/eval_odinw.py --config_file configs/pretrain/mq-glip-t.yaml \
|
||||
--opts 'MODEL.WEIGHT ${model_weight_path}' \
|
||||
--setting finetuning-free \
|
||||
--add_name tiny \
|
||||
--log_path 'OUTPUT/odinw_log/'
|
||||
```
|
||||
The results are stored at ``OUTPUT/odinw_log/``.
|
||||
|
||||
If you wish to use custom vision queries or datasets, specify ``--task_config`` and ``--custom_bank_path``. The ``task_config`` should be like the ones in [ODinW configs](configs/odinw_13/AerialMaritimeDrone_large.yaml). The ``custom_bank_path`` should be extracted following the [instruction](#vision-query-extraction). For example,
|
||||
```
|
||||
python tools/eval_odinw.py --config_file configs/pretrain/mq-glip-t.yaml \
|
||||
--opts 'MODEL.WEIGHT ${model_weight_path}' \
|
||||
--setting finetuning-free \
|
||||
--add_name tiny \
|
||||
--log_path 'OUTPUT/custom_log/'
|
||||
--task_config ${custom_config_path}
|
||||
--custom_bank_path ${custom_bank_path}
|
||||
```
|
||||
|
||||
|
||||
## Fine-Tuning
|
||||
**Take MQ-GLIP-T as an example.**
|
||||
```
|
||||
python tools/eval_odinw.py --config_file configs/pretrain/mq-glip-t.yaml \
|
||||
--opts 'MODEL.WEIGHT ${model_weight_path}' \
|
||||
--setting 3-shot \
|
||||
--add_name tiny \
|
||||
--log_path 'OUTPUT/odinw_log/'
|
||||
```
|
||||
This command will first automatically extract the vision query bank from the (few-shot) training set. Then conduct fine-tuning.
|
||||
If you wish to use custom vision queries, add ``'VISION_QUERY.QUERY_BANK_PATH custom_bank_path'`` to the ``--opts`` argment, and also modify the ``dataset_configs`` in the ``tools/eval_odinw.py``.
|
||||
|
||||
If set ``VISION_QUERY.QUERY_BANK_PATH`` to ``''``, the model will automatically extract the vision query bank from the (few-shot) training set before fine-tuning.
|
||||
|
||||
|
||||
## Single-Modal Evaluation
|
||||
|
||||
Here we provide introduction on utilizing single modal queries, such as visual exemplars or textual description.
|
||||
|
||||
|
||||
Follow the command as in [``Finetuning-free Evaluation``](#finetuning-free-evaluation). But set the following hyper-parameters.
|
||||
|
||||
To solely use vision queries, add hyper-parameters:
|
||||
```
|
||||
VISION_QUERY.MASK_DURING_INFERENCE True VISION_QUERY.TEXT_DROPOUT 1.0
|
||||
```
|
||||
|
||||
To solely use language queries, add hyper-parameters:
|
||||
```
|
||||
VISION_QUERY.ENABLED FALSE
|
||||
```
|
||||
|
||||
For example, to solely use vision queries,
|
||||
```
|
||||
python -m torch.distributed.launch --nproc_per_node=4 \
|
||||
tools/test_grounding_net.py \
|
||||
--config-file configs/pretrain/mq-glip-t.yaml \
|
||||
--additional_model_config configs/vision_query_5shot/lvis_minival.yaml \
|
||||
VISION_QUERY.QUERY_BANK_PATH MODEL/lvis_query_5_pool7_sel_tiny.pth \
|
||||
MODEL.WEIGHT ${model_weight_path} \
|
||||
TEST.IMS_PER_BATCH 4 \
|
||||
VISION_QUERY.MASK_DURING_INFERENCE True VISION_QUERY.TEXT_DROPOUT 1.0
|
||||
```
|
||||
|
||||
```
|
||||
python tools/eval_odinw.py --config_file configs/pretrain/mq-glip-t.yaml \
|
||||
--opts 'MODEL.WEIGHT ${model_weight_path} VISION_QUERY.MASK_DURING_INFERENCE True VISION_QUERY.TEXT_DROPOUT 1.0' \
|
||||
--setting finetuning-free \
|
||||
--add_name tiny \
|
||||
--log_path 'OUTPUT/odinw_log/'
|
||||
```
|
||||
|
|
@ -0,0 +1,116 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "boat", "supercategory": "movable-objects"},
|
||||
{"id": 2, "name": "car", "supercategory": "movable-objects"}, {"id": 3, "name":
|
||||
"dock", "supercategory": "movable-objects"}, {"id": 4, "name": "jetski", "supercategory":
|
||||
"movable-objects"}, {"id": 5, "name": "lift", "supercategory": "movable-objects"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/test/annotations_without_background.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/test
|
||||
train:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/annotations_without_background.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_10_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_10_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_10_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_1_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_1_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_1_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_3_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_3_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_3_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_5_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_5_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
train_5_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train
|
||||
val:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/annotations_without_background.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 6
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 6
|
||||
FCOS:
|
||||
NUM_CLASSES: 6
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 6
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'AerialDrone'
|
||||
|
|
@ -0,0 +1,123 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "fish", "supercategory": "creatures"}, {"id":
|
||||
2, "name": "jellyfish", "supercategory": "creatures"}, {"id": 3, "name": "penguin",
|
||||
"supercategory": "creatures"}, {"id": 4, "name": "puffin", "supercategory": "creatures"},
|
||||
{"id": 5, "name": "shark", "supercategory": "creatures"}, {"id": 6, "name": "starfish",
|
||||
"supercategory": "creatures"}, {"id": 7, "name": "stingray", "supercategory":
|
||||
"creatures"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test/annotations_without_background.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test
|
||||
train:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/annotations_without_background.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_wrong:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/annotations_without_background_wrong_label.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_10_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_10_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_10_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_1_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_1_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_1_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_3_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_3_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_3_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_5_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_5_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
train_5_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train
|
||||
val:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/annotations_without_background.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Aquarium'
|
||||
GROUNDINGDINO:
|
||||
box_threshold: 0.08
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Cottontail-Rabbit", "supercategory": "Cottontail-Rabbit"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/CottontailRabbits/test/annotations_without_background.json
|
||||
img_dir: odinw/CottontailRabbits/test
|
||||
train:
|
||||
ann_file: odinw/CottontailRabbits/train/annotations_without_background.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_10_3:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_10_30:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_10_300:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_1_3:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_1_30:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_1_300:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_3_3:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_3_30:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_3_300:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_5_3:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_5_30:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
train_5_300:
|
||||
ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/train
|
||||
val:
|
||||
ann_file: odinw/CottontailRabbits/valid/annotations_without_background.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/CottontailRabbits/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 2
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 2
|
||||
FCOS:
|
||||
NUM_CLASSES: 2
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 2
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Rabbits'
|
||||
|
|
@ -0,0 +1,152 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "hand", "supercategory": "hands"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/annotations_without_background.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_10_3:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_10_30:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_10_300:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_1_3:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_1_30:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_1_300:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_3_3:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_3_30:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_3_300:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_5_3:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_5_30:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
minival_5_300:
|
||||
ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/mini_val
|
||||
test:
|
||||
ann_file: odinw/EgoHands/generic/test/annotations_without_background.json
|
||||
img_dir: odinw/EgoHands/generic/test
|
||||
train:
|
||||
ann_file: odinw/EgoHands/generic/train/annotations_without_background.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_10_3:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_10_30:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_10_300:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_1_3:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_1_30:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_1_300:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_3_3:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_3_30:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_3_300:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_5_3:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_5_30:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
train_5_300:
|
||||
ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/train
|
||||
val:
|
||||
ann_file: odinw/EgoHands/generic/valid/annotations_without_background.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/EgoHands/generic/valid
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 2
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 2
|
||||
FCOS:
|
||||
NUM_CLASSES: 2
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 2
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'EgoHands'
|
||||
|
|
@ -0,0 +1,119 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "CoW", "supercategory": "mushroom"}, {"id":
|
||||
2, "name": "chanterelle", "supercategory": "mushroom"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test/annotations_without_background.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test
|
||||
train:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/annotations_without_background.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_wrong:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/annotations_without_background_wrong_label.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_10_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_10_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_10_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_1_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_1_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_1_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_3_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_3_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_3_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_5_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_5_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
train_5_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train
|
||||
val:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/annotations_without_background.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 3
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 3
|
||||
FCOS:
|
||||
NUM_CLASSES: 3
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 3
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Mushrooms'
|
||||
GROUNDINGDINO:
|
||||
box_threshold: 0.08
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "package", "supercategory": "packages"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/Packages/Raw/test/annotations_without_background.json
|
||||
img_dir: odinw/Packages/Raw/test
|
||||
train:
|
||||
ann_file: odinw/Packages/Raw/train/annotations_without_background.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_10_3:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_10_30:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_10_300:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_1_3:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_1_30:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_1_300:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_3_3:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_3_30:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_3_300:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_5_3:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_5_30:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
train_5_300:
|
||||
ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/Packages/Raw/train
|
||||
val:
|
||||
ann_file: odinw/Packages/Raw/valid/annotations_without_background.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/Packages/Raw/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 2
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 2
|
||||
FCOS:
|
||||
NUM_CLASSES: 2
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 2
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Packages'
|
||||
|
|
@ -0,0 +1,126 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "aeroplane", "supercategory": "VOC"}, {"id":
|
||||
2, "name": "bicycle", "supercategory": "VOC"}, {"id": 3, "name": "bird", "supercategory":
|
||||
"VOC"}, {"id": 4, "name": "boat", "supercategory": "VOC"}, {"id": 5, "name": "bottle",
|
||||
"supercategory": "VOC"}, {"id": 6, "name": "bus", "supercategory": "VOC"}, {"id":
|
||||
7, "name": "car", "supercategory": "VOC"}, {"id": 8, "name": "cat", "supercategory":
|
||||
"VOC"}, {"id": 9, "name": "chair", "supercategory": "VOC"}, {"id": 10, "name":
|
||||
"cow", "supercategory": "VOC"}, {"id": 11, "name": "diningtable", "supercategory":
|
||||
"VOC"}, {"id": 12, "name": "dog", "supercategory": "VOC"}, {"id": 13, "name":
|
||||
"horse", "supercategory": "VOC"}, {"id": 14, "name": "motorbike", "supercategory":
|
||||
"VOC"}, {"id": 15, "name": "person", "supercategory": "VOC"}, {"id": 16, "name":
|
||||
"pottedplant", "supercategory": "VOC"}, {"id": 17, "name": "sheep", "supercategory":
|
||||
"VOC"}, {"id": 18, "name": "sofa", "supercategory": "VOC"}, {"id": 19, "name":
|
||||
"train", "supercategory": "VOC"}, {"id": 20, "name": "tvmonitor", "supercategory":
|
||||
"VOC"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/PascalVOC/valid/annotations_without_background.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
train:
|
||||
ann_file: odinw/PascalVOC/train/annotations_without_background.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_10_3:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_10_30:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_10_300:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_1_3:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_1_30:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_1_300:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_3_3:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_3_30:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_3_300:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_5_3:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_5_30:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
train_5_300:
|
||||
ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/PascalVOC/train
|
||||
val:
|
||||
ann_file: odinw/PascalVOC/valid/annotations_without_background.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/PascalVOC/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 21
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 21
|
||||
FCOS:
|
||||
NUM_CLASSES: 21
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 21
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'PascalVOC'
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "raccoon", "supercategory": "raccoons"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/test/annotations_without_background.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/test
|
||||
train:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/annotations_without_background.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_10_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_10_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_10_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_1_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_1_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_1_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_3_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_3_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_3_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_5_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_5_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
train_5_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train
|
||||
val:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/annotations_without_background.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 2
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 2
|
||||
FCOS:
|
||||
NUM_CLASSES: 2
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 2
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Raccoon'
|
||||
|
|
@ -0,0 +1,115 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Crab", "supercategory": "shellfish"}, {"id":
|
||||
2, "name": "Lobster", "supercategory": "shellfish"}, {"id": 3, "name": "Shrimp",
|
||||
"supercategory": "shellfish"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/test/annotations_without_background.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/test
|
||||
train:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/annotations_without_background.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_10_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_10_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_10_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_1_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_1_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_1_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_3_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_3_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_3_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_5_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_5_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
train_5_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train
|
||||
val:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/annotations_without_background.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 4
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 4
|
||||
FCOS:
|
||||
NUM_CLASSES: 4
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 4
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Shellfish'
|
||||
|
|
@ -0,0 +1,155 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Ambulance", "supercategory": "vehicles"},
|
||||
{"id": 2, "name": "Bus", "supercategory": "vehicles"}, {"id": 3, "name": "Car",
|
||||
"supercategory": "vehicles"}, {"id": 4, "name": "Motorcycle", "supercategory":
|
||||
"vehicles"}, {"id": 5, "name": "Truck", "supercategory": "vehicles"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/annotations_without_background.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_10_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_10_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_10_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_1_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_1_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_1_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_3_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_3_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_3_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_5_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_5_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
minival_5_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val
|
||||
test:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/test/annotations_without_background.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/test
|
||||
train:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/annotations_without_background.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_10_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_10_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_10_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_1_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_1_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_1_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_3_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_3_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_3_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_5_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_5_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
train_5_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train
|
||||
val:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/annotations_without_background.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 6
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 6
|
||||
FCOS:
|
||||
NUM_CLASSES: 6
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 6
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Vehicles'
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "pistol", "supercategory": "Guns"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/pistols/export/test_annotations_without_background.json
|
||||
img_dir: odinw/pistols/export
|
||||
train:
|
||||
ann_file: odinw/pistols/export/train_annotations_without_background.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_10_3:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_10_30:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_10_300:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_1_3:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_1_30:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_1_300:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_3_3:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_3_30:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_3_300:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_5_3:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_5_30:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
train_5_300:
|
||||
ann_file: odinw/pistols/export/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
val:
|
||||
ann_file: odinw/pistols/export/val_annotations_without_background.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_10_3:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_10_30:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_10_300:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_1_3:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_1_30:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_1_300:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_3_3:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_3_30:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_3_300:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_5_3:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_5_30:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/pistols/export
|
||||
val_5_300:
|
||||
ann_file: odinw/pistols/export/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/pistols/export
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 297
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 297
|
||||
FCOS:
|
||||
NUM_CLASSES: 297
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 297
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Pistols'
|
||||
|
|
@ -0,0 +1,113 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "pothole", "supercategory": "potholes"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/pothole/test/annotations_without_background.json
|
||||
img_dir: odinw/pothole/test
|
||||
train:
|
||||
ann_file: odinw/pothole/train/annotations_without_background.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_10_3:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_10_30:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_10_300:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_1_3:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_1_30:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_1_300:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_3_3:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_3_30:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_3_300:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_5_3:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_5_30:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/pothole/train
|
||||
train_5_300:
|
||||
ann_file: odinw/pothole/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/pothole/train
|
||||
val:
|
||||
ann_file: odinw/pothole/valid/annotations_without_background.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/pothole/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/pothole/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/pothole/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 2
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 2
|
||||
FCOS:
|
||||
NUM_CLASSES: 2
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 2
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Pothole'
|
||||
|
|
@ -0,0 +1,114 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "dog", "supercategory": "dogs-person"}, {"id":
|
||||
2, "name": "person", "supercategory": "dogs-person"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/thermalDogsAndPeople/test/annotations_without_background.json
|
||||
img_dir: odinw/thermalDogsAndPeople/test
|
||||
train:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/annotations_without_background.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_10_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_10_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_10_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_1_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_1_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_1_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_3_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_3_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_3_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_5_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_5_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
train_5_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/train
|
||||
val:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/annotations_without_background.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/thermalDogsAndPeople/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 3
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 3
|
||||
FCOS:
|
||||
NUM_CLASSES: 3
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 3
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Thermal'
|
||||
|
|
@ -0,0 +1,76 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "boat", "supercategory": "movable-objects"},
|
||||
{"id": 2, "name": "car", "supercategory": "movable-objects"}, {"id": 3, "name":
|
||||
"dock", "supercategory": "movable-objects"}, {"id": 4, "name": "jetski", "supercategory":
|
||||
"movable-objects"}, {"id": 5, "name": "lift", "supercategory": "movable-objects"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/AerialMaritimeDrone/large/test/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/test}
|
||||
train: {ann_file: odinw/AerialMaritimeDrone/large/train/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_10_3: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_10_30: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_10_300: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_1_3: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_1_30: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_1_300: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_3_3: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_3_30: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_3_300: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_5_3: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_5_30: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
train_5_300: {ann_file: odinw/AerialMaritimeDrone/large/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/train}
|
||||
val: {ann_file: odinw/AerialMaritimeDrone/large/valid/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_10_3: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_10_30: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_10_300: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_1_3: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_1_30: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_1_300: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_3_3: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_3_30: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_3_300: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_5_3: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_5_30: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
val_5_300: {ann_file: odinw/AerialMaritimeDrone/large/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/large/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 6}
|
||||
DYHEAD: {NUM_CLASSES: 6}
|
||||
FCOS: {NUM_CLASSES: 6}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 6}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'AerialMaritimeDrone_large'
|
||||
|
|
@ -0,0 +1,76 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "boat", "supercategory": "movable-objects"},
|
||||
{"id": 2, "name": "car", "supercategory": "movable-objects"}, {"id": 3, "name":
|
||||
"dock", "supercategory": "movable-objects"}, {"id": 4, "name": "jetski", "supercategory":
|
||||
"movable-objects"}, {"id": 5, "name": "lift", "supercategory": "movable-objects"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/AerialMaritimeDrone/tiled/test/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/test}
|
||||
train: {ann_file: odinw/AerialMaritimeDrone/tiled/train/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_10_3: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_10_30: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_10_300: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_1_3: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_1_30: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_1_300: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_3_3: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_3_30: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_3_300: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_5_3: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_5_30: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
train_5_300: {ann_file: odinw/AerialMaritimeDrone/tiled/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/train}
|
||||
val: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/annotations_without_background.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_10_3: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_10_30: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_10_300: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_1_3: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_1_30: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_1_300: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_3_3: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_3_30: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_3_300: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_5_3: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_5_30: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
val_5_300: {ann_file: odinw/AerialMaritimeDrone/tiled/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/AerialMaritimeDrone/tiled/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 6}
|
||||
DYHEAD: {NUM_CLASSES: 6}
|
||||
FCOS: {NUM_CLASSES: 6}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 6}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'AerialMaritimeDrone_tiled'
|
||||
|
|
@ -0,0 +1,108 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "A", "supercategory": "Letters"}, {"id":
|
||||
2, "name": "B", "supercategory": "Letters"}, {"id": 3, "name": "C", "supercategory":
|
||||
"Letters"}, {"id": 4, "name": "D", "supercategory": "Letters"}, {"id": 5, "name":
|
||||
"E", "supercategory": "Letters"}, {"id": 6, "name": "F", "supercategory": "Letters"},
|
||||
{"id": 7, "name": "G", "supercategory": "Letters"}, {"id": 8, "name": "H", "supercategory":
|
||||
"Letters"}, {"id": 9, "name": "I", "supercategory": "Letters"}, {"id": 10, "name":
|
||||
"J", "supercategory": "Letters"}, {"id": 11, "name": "K", "supercategory": "Letters"},
|
||||
{"id": 12, "name": "L", "supercategory": "Letters"}, {"id": 13, "name": "M", "supercategory":
|
||||
"Letters"}, {"id": 14, "name": "N", "supercategory": "Letters"}, {"id": 15, "name":
|
||||
"O", "supercategory": "Letters"}, {"id": 16, "name": "P", "supercategory": "Letters"},
|
||||
{"id": 17, "name": "Q", "supercategory": "Letters"}, {"id": 18, "name": "R", "supercategory":
|
||||
"Letters"}, {"id": 19, "name": "S", "supercategory": "Letters"}, {"id": 20, "name":
|
||||
"T", "supercategory": "Letters"}, {"id": 21, "name": "U", "supercategory": "Letters"},
|
||||
{"id": 22, "name": "V", "supercategory": "Letters"}, {"id": 23, "name": "W", "supercategory":
|
||||
"Letters"}, {"id": 24, "name": "X", "supercategory": "Letters"}, {"id": 25, "name":
|
||||
"Y", "supercategory": "Letters"}, {"id": 26, "name": "Z", "supercategory": "Letters"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/test}
|
||||
train: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/train}
|
||||
train_10_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot10_seed3.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_10_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot10_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_10_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot10_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_1_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot1_seed3.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_1_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot1_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_1_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot1_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_3_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot3_seed3.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_3_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot3_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_3_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot3_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_5_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot5_seed3.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_5_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot5_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
train_5_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/train/fewshot_train_shot5_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/train}
|
||||
val: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid}
|
||||
val_10_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_10_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot10_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_10_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot10_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_1_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid}
|
||||
val_1_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_1_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot1_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_3_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid}
|
||||
val_3_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_3_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot3_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_5_3: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/AmericanSignLanguageLetters/American Sign Language Letters.v1-v1.coco/valid}
|
||||
val_5_30: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
val_5_300: {ann_file: odinw/AmericanSignLanguageLetters/American Sign Language
|
||||
Letters.v1-v1.coco/valid/fewshot_val_shot5_seed300.json, img_dir: odinw/AmericanSignLanguageLetters/American
|
||||
Sign Language Letters.v1-v1.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 27}
|
||||
DYHEAD: {NUM_CLASSES: 27}
|
||||
FCOS: {NUM_CLASSES: 27}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 27}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'AmericanSignLanguageLetters'
|
||||
|
|
@ -0,0 +1,84 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
CAPTION_PROMPT: '[{"prefix": " ", "name": "fish", "suffix": ""}, {"prefix": "",
|
||||
"name": "jellyfish", "suffix": ""}, {"prefix": "", "name": "penguin", "suffix":
|
||||
" , which is black and white"}, {"prefix": "", "name": "puffin", "suffix": " with
|
||||
orange beaks "}, {"prefix": "", "name": "shark", "suffix": ""}, {"prefix": "",
|
||||
"name": "starfish", "suffix": ""}, {"prefix": "", "name": "stingray", "suffix":
|
||||
" which is flat and round"}, ]'
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "fish", "supercategory": "creatures"}, {"id":
|
||||
2, "name": "jellyfish", "supercategory": "creatures"}, {"id": 3, "name": "penguin",
|
||||
"supercategory": "creatures"}, {"id": 4, "name": "puffin", "supercategory": "creatures"},
|
||||
{"id": 5, "name": "shark", "supercategory": "creatures"}, {"id": 6, "name": "starfish",
|
||||
"supercategory": "creatures"}, {"id": 7, "name": "stingray", "supercategory":
|
||||
"creatures"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/test}
|
||||
train: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_10_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_10_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_10_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_1_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_1_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_1_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_3_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_3_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_3_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_5_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_5_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
train_5_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/train}
|
||||
val: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_10_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_10_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_10_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_1_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_1_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_1_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_3_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_3_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_3_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_5_3: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_5_30: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
val_5_300: {ann_file: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/Aquarium/Aquarium Combined.v2-raw-1024.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 8}
|
||||
DYHEAD: {NUM_CLASSES: 8}
|
||||
FCOS: {NUM_CLASSES: 8}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 8}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Aquarium'
|
||||
|
||||
|
|
@ -0,0 +1,74 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Platelets", "supercategory": "cells"}, {"id":
|
||||
2, "name": "RBC", "supercategory": "cells"}, {"id": 3, "name": "WBC", "supercategory":
|
||||
"cells"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/test}
|
||||
train: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_10_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_10_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_10_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_1_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_1_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_1_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_3_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_3_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_3_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_5_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_5_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
train_5_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/train}
|
||||
val: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_10_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_10_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_10_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_1_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_1_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_1_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_3_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_3_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_3_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_5_3: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_5_30: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
val_5_300: {ann_file: odinw/BCCD/BCCD.v3-raw.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/BCCD/BCCD.v3-raw.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 4}
|
||||
DYHEAD: {NUM_CLASSES: 4}
|
||||
FCOS: {NUM_CLASSES: 4}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 4}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'BCCD_BCCD'
|
||||
|
|
@ -0,0 +1,81 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": " ", "supercategory": "pieces"}, {"id":
|
||||
2, "name": "black bishop", "supercategory": "pieces"}, {"id": 3, "name": "black
|
||||
king", "supercategory": "pieces"}, {"id": 4, "name": "black knight", "supercategory":
|
||||
"pieces"}, {"id": 5, "name": "black pawn", "supercategory": "pieces"}, {"id":
|
||||
6, "name": "black queen", "supercategory": "pieces"}, {"id": 7, "name": "black
|
||||
rook", "supercategory": "pieces"}, {"id": 8, "name": "white bishop", "supercategory":
|
||||
"pieces"}, {"id": 9, "name": "white king", "supercategory": "pieces"}, {"id":
|
||||
10, "name": "white knight", "supercategory": "pieces"}, {"id": 11, "name": "white
|
||||
pawn", "supercategory": "pieces"}, {"id": 12, "name": "white queen", "supercategory":
|
||||
"pieces"}, {"id": 13, "name": "white rook", "supercategory": "pieces"}]'
|
||||
PREDEFINED_TEXT: odinw/original/ChessPieces/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/test}
|
||||
train: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_10_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_10_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_10_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_1_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_1_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_1_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_3_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_3_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_3_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_5_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_5_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
train_5_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/train}
|
||||
val: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_10_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_10_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_10_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_1_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_1_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_1_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_3_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_3_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_3_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_5_3: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_5_30: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
val_5_300: {ann_file: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/ChessPieces/Chess Pieces.v23-raw.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 14}
|
||||
DYHEAD: {NUM_CLASSES: 14}
|
||||
FCOS: {NUM_CLASSES: 14}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 14}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'ChessPiece'
|
||||
|
|
@ -0,0 +1,72 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "rabbit", "supercategory": "Cottontail-Rabbit"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/CottontailRabbits/test/annotations_without_background.json,
|
||||
img_dir: odinw/CottontailRabbits/test}
|
||||
train: {ann_file: odinw/CottontailRabbits/train/annotations_without_background.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_10_3: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_10_30: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_10_300: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_1_3: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_1_30: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_1_300: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_3_3: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_3_30: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_3_300: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_5_3: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_5_30: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
train_5_300: {ann_file: odinw/CottontailRabbits/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/train}
|
||||
val: {ann_file: odinw/CottontailRabbits/valid/annotations_without_background.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_10_3: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_10_30: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_10_300: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_1_3: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_1_30: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_1_300: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_3_3: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_3_30: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_3_300: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_5_3: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_5_30: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
val_5_300: {ann_file: odinw/CottontailRabbits/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/CottontailRabbits/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'CottontailRabbits'
|
||||
|
|
@ -0,0 +1,103 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 8
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "follow", "supercategory": "actions"}, {"id":
|
||||
2, "name": "follow_hand", "supercategory": "actions"}, {"id": 3, "name": "land",
|
||||
"supercategory": "actions"}, {"id": 4, "name": "land_hand", "supercategory": "actions"},
|
||||
{"id": 5, "name": "null", "supercategory": "actions"}, {"id": 6, "name": "object",
|
||||
"supercategory": "actions"}, {"id": 7, "name": "takeoff", "supercategory": "actions"},
|
||||
{"id": 8, "name": "takeoff-hand", "supercategory": "actions"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_10_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_10_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_10_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_1_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_1_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_1_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_3_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_3_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_3_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_5_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_5_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
minival_5_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/mini_val}
|
||||
test: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/test}
|
||||
train: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_10_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_10_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_10_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_1_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_1_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_1_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_3_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_3_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_3_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_5_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_5_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
train_5_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/train}
|
||||
val: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_10_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_10_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_10_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_1_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_1_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_1_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_3_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_3_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_3_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_5_3: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_5_30: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
val_5_300: {ann_file: odinw/DroneControl/Drone Control.v3-raw.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/DroneControl/Drone Control.v3-raw.coco/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 9}
|
||||
DYHEAD: {NUM_CLASSES: 9}
|
||||
FCOS: {NUM_CLASSES: 9}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 9}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'DroneControl_Drone_Control'
|
||||
|
|
@ -0,0 +1,99 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
CAPTION_PROMPT: '[{"prefix": " ", "name": "hand", "suffix": " of a person"},]'
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "hand", "supercategory": "hands"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/EgoHands/generic/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_10_3: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_10_30: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_10_300: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_1_3: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_1_30: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_1_300: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_3_3: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_3_30: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_3_300: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_5_3: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_5_30: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
minival_5_300: {ann_file: odinw/EgoHands/generic/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/mini_val}
|
||||
test: {ann_file: odinw/EgoHands/generic/test/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/generic/test}
|
||||
train: {ann_file: odinw/EgoHands/generic/train/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_10_3: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_10_30: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_10_300: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_1_3: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_1_30: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_1_300: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_3_3: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_3_30: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_3_300: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_5_3: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_5_30: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
train_5_300: {ann_file: odinw/EgoHands/generic/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/train}
|
||||
val: {ann_file: odinw/EgoHands/generic/valid/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_10_3: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_10_30: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_10_300: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_1_3: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_1_30: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_1_300: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_3_3: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_3_30: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_3_300: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_5_3: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_5_30: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
val_5_300: {ann_file: odinw/EgoHands/generic/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/generic/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'EgoHands_generic'
|
||||
|
|
@ -0,0 +1,100 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "myleft", "supercategory": "hands"}, {"id":
|
||||
2, "name": "myright", "supercategory": "hands"}, {"id": 3, "name": "yourleft",
|
||||
"supercategory": "hands"}, {"id": 4, "name": "yourright", "supercategory": "hands"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/EgoHands/specific/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_10_3: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_10_30: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_10_300: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_1_3: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_1_30: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_1_300: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_3_3: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_3_30: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_3_300: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_5_3: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_5_30: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
minival_5_300: {ann_file: odinw/EgoHands/specific/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/mini_val}
|
||||
test: {ann_file: odinw/EgoHands/specific/test/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/specific/test}
|
||||
train: {ann_file: odinw/EgoHands/specific/train/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_10_3: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_10_30: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_10_300: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_1_3: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_1_30: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_1_300: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_3_3: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_3_30: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_3_300: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_5_3: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_5_30: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
train_5_300: {ann_file: odinw/EgoHands/specific/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/train}
|
||||
val: {ann_file: odinw/EgoHands/specific/valid/annotations_without_background.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_10_3: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_10_30: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_10_300: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_1_3: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_1_30: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_1_300: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_3_3: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_3_30: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_3_300: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_5_3: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_5_30: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
val_5_300: {ann_file: odinw/EgoHands/specific/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/EgoHands/specific/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 5}
|
||||
DYHEAD: {NUM_CLASSES: 5}
|
||||
FCOS: {NUM_CLASSES: 5}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 5}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'EgoHands_specific'
|
||||
|
|
@ -0,0 +1,74 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "head", "supercategory": "Workers"}, {"id":
|
||||
2, "name": "helmet", "supercategory": "Workers"}, {"id": 3, "name": "person",
|
||||
"supercategory": "Workers"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/HardHatWorkers/raw/test/annotations_without_background.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/test}
|
||||
train: {ann_file: odinw/HardHatWorkers/raw/train/annotations_without_background.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_10_3: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_10_30: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_10_300: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_1_3: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_1_30: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_1_300: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_3_3: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_3_30: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_3_300: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_5_3: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_5_30: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
train_5_300: {ann_file: odinw/HardHatWorkers/raw/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/train}
|
||||
val: {ann_file: odinw/HardHatWorkers/raw/valid/annotations_without_background.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_10_3: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_10_30: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_10_300: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_1_3: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_1_30: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_1_300: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_3_3: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_3_30: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_3_300: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_5_3: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_5_30: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
val_5_300: {ann_file: odinw/HardHatWorkers/raw/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/HardHatWorkers/raw/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 4}
|
||||
DYHEAD: {NUM_CLASSES: 4}
|
||||
FCOS: {NUM_CLASSES: 4}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 4}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'HardHatWorkers'
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "mask", "supercategory": "People"}, {"id":
|
||||
2, "name": "no-mask", "supercategory": "People"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/MaskWearing/raw/test/annotations_without_background.json,
|
||||
img_dir: odinw/MaskWearing/raw/test}
|
||||
train: {ann_file: odinw/MaskWearing/raw/train/annotations_without_background.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_10_3: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_10_30: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_10_300: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_1_3: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_1_30: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_1_300: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_3_3: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_3_30: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_3_300: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_5_3: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_5_30: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
train_5_300: {ann_file: odinw/MaskWearing/raw/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/train}
|
||||
val: {ann_file: odinw/MaskWearing/raw/valid/annotations_without_background.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_10_3: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_10_30: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_10_300: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_1_3: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_1_30: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_1_300: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_3_3: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_3_30: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_3_300: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_5_3: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_5_30: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
val_5_300: {ann_file: odinw/MaskWearing/raw/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/MaskWearing/raw/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'MaskWearing'
|
||||
|
|
@ -0,0 +1,72 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "bottle", "supercategory": "bottles"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/MountainDewCommercial/test/annotations_without_background.json,
|
||||
img_dir: odinw/MountainDewCommercial/test}
|
||||
train: {ann_file: odinw/MountainDewCommercial/train/annotations_without_background.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_10_3: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_10_30: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_10_300: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_1_3: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_1_30: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_1_300: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_3_3: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_3_30: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_3_300: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_5_3: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_5_30: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
train_5_300: {ann_file: odinw/MountainDewCommercial/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/train}
|
||||
val: {ann_file: odinw/MountainDewCommercial/valid/annotations_without_background.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_10_3: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_10_30: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_10_300: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_1_3: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_1_30: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_1_300: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_3_3: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_3_30: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_3_300: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_5_3: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_5_30: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
val_5_300: {ann_file: odinw/MountainDewCommercial/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/MountainDewCommercial/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'MountainDewCommercial'
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{''id'': 1, ''name'': ''flat mushroom'', ''supercategory'':
|
||||
''mushroom''}, {''id'': 2, ''name'': ''yellow mushroom'', ''supercategory'': ''mushroom''}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/test}
|
||||
train: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_10_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_10_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_10_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_1_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_1_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_1_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_3_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_3_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_3_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_5_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_5_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
train_5_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/train}
|
||||
val: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_10_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_10_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_10_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_1_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_1_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_1_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_3_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_3_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_3_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_5_3: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_5_30: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
val_5_300: {ann_file: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/NorthAmericaMushrooms/North American Mushrooms.v1-416x416.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'NorthAmericaMushrooms'
|
||||
|
|
@ -0,0 +1,126 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "cat-Abyssinian", "supercategory": "pets"},
|
||||
{"id": 2, "name": "cat-Bengal", "supercategory": "pets"}, {"id": 3, "name": "cat-Birman",
|
||||
"supercategory": "pets"}, {"id": 4, "name": "cat-Bombay", "supercategory": "pets"},
|
||||
{"id": 5, "name": "cat-British_Shorthair", "supercategory": "pets"}, {"id": 6,
|
||||
"name": "cat-Egyptian_Mau", "supercategory": "pets"}, {"id": 7, "name": "cat-Maine_Coon",
|
||||
"supercategory": "pets"}, {"id": 8, "name": "cat-Persian", "supercategory": "pets"},
|
||||
{"id": 9, "name": "cat-Ragdoll", "supercategory": "pets"}, {"id": 10, "name":
|
||||
"cat-Russian_Blue", "supercategory": "pets"}, {"id": 11, "name": "cat-Siamese",
|
||||
"supercategory": "pets"}, {"id": 12, "name": "cat-Sphynx", "supercategory": "pets"},
|
||||
{"id": 13, "name": "dog-american_bulldog", "supercategory": "pets"}, {"id": 14,
|
||||
"name": "dog-american_pit_bull_terrier", "supercategory": "pets"}, {"id": 15,
|
||||
"name": "dog-basset_hound", "supercategory": "pets"}, {"id": 16, "name": "dog-beagle",
|
||||
"supercategory": "pets"}, {"id": 17, "name": "dog-boxer", "supercategory": "pets"},
|
||||
{"id": 18, "name": "dog-chihuahua", "supercategory": "pets"}, {"id": 19, "name":
|
||||
"dog-english_cocker_spaniel", "supercategory": "pets"}, {"id": 20, "name": "dog-english_setter",
|
||||
"supercategory": "pets"}, {"id": 21, "name": "dog-german_shorthaired", "supercategory":
|
||||
"pets"}, {"id": 22, "name": "dog-great_pyrenees", "supercategory": "pets"}, {"id":
|
||||
23, "name": "dog-havanese", "supercategory": "pets"}, {"id": 24, "name": "dog-japanese_chin",
|
||||
"supercategory": "pets"}, {"id": 25, "name": "dog-keeshond", "supercategory":
|
||||
"pets"}, {"id": 26, "name": "dog-leonberger", "supercategory": "pets"}, {"id":
|
||||
27, "name": "dog-miniature_pinscher", "supercategory": "pets"}, {"id": 28, "name":
|
||||
"dog-newfoundland", "supercategory": "pets"}, {"id": 29, "name": "dog-pomeranian",
|
||||
"supercategory": "pets"}, {"id": 30, "name": "dog-pug", "supercategory": "pets"},
|
||||
{"id": 31, "name": "dog-saint_bernard", "supercategory": "pets"}, {"id": 32, "name":
|
||||
"dog-samoyed", "supercategory": "pets"}, {"id": 33, "name": "dog-scottish_terrier",
|
||||
"supercategory": "pets"}, {"id": 34, "name": "dog-shiba_inu", "supercategory":
|
||||
"pets"}, {"id": 35, "name": "dog-staffordshire_bull_terrier", "supercategory":
|
||||
"pets"}, {"id": 36, "name": "dog-wheaten_terrier", "supercategory": "pets"}, {"id":
|
||||
37, "name": "dog-yorkshire_terrier", "supercategory": "pets"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/OxfordPets/by-breed/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_10_3: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_10_30: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_10_300: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_1_3: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_1_30: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_1_300: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_3_3: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_3_30: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_3_300: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_5_3: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_5_30: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
minival_5_300: {ann_file: odinw/OxfordPets/by-breed/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/mini_val}
|
||||
test: {ann_file: odinw/OxfordPets/by-breed/test/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/test}
|
||||
train: {ann_file: odinw/OxfordPets/by-breed/train/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_10_3: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_10_30: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_10_300: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_1_3: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_1_30: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_1_300: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_3_3: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_3_30: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_3_300: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_5_3: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_5_30: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
train_5_300: {ann_file: odinw/OxfordPets/by-breed/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/train}
|
||||
val: {ann_file: odinw/OxfordPets/by-breed/valid/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_10_3: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_10_30: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_10_300: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_1_3: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_1_30: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_1_300: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_3_3: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_3_30: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_3_300: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_5_3: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_5_30: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
val_5_300: {ann_file: odinw/OxfordPets/by-breed/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-breed/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 38}
|
||||
DYHEAD: {NUM_CLASSES: 38}
|
||||
FCOS: {NUM_CLASSES: 38}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 38}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'OxfordPets_by-breed'
|
||||
|
|
@ -0,0 +1,99 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "cat", "supercategory": "pets"}, {"id": 2,
|
||||
"name": "dog", "supercategory": "pets"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/OxfordPets/by-species/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_10_3: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_10_30: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_10_300: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_1_3: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_1_30: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_1_300: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_3_3: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_3_30: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_3_300: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_5_3: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_5_30: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
minival_5_300: {ann_file: odinw/OxfordPets/by-species/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/mini_val}
|
||||
test: {ann_file: odinw/OxfordPets/by-species/test/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-species/test}
|
||||
train: {ann_file: odinw/OxfordPets/by-species/train/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_10_3: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_10_30: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_10_300: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_1_3: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_1_30: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_1_300: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_3_3: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_3_30: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_3_300: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_5_3: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_5_30: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
train_5_300: {ann_file: odinw/OxfordPets/by-species/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/train}
|
||||
val: {ann_file: odinw/OxfordPets/by-species/valid/annotations_without_background.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_10_3: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_10_30: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_10_300: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_1_3: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_1_30: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_1_300: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_3_3: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_3_30: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_3_300: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_5_3: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_5_30: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
val_5_300: {ann_file: odinw/OxfordPets/by-species/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/OxfordPets/by-species/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'OxfordPets_by-species'
|
||||
|
|
@ -0,0 +1,82 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "space-empty", "supercategory": "spaces"},
|
||||
{"id": 2, "name": "space-occupied", "supercategory": "spaces"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/PKLot/640/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_10_3: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_10_30: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_10_300: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_1_3: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_1_30: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_1_300: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_3_3: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_3_30: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_3_300: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_5_3: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_5_30: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
minival_5_300: {ann_file: odinw/PKLot/640/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/PKLot/640/mini_val}
|
||||
test: {ann_file: odinw/PKLot/640/test/annotations_without_background.json, img_dir: odinw/PKLot/640/test}
|
||||
train: {ann_file: odinw/PKLot/640/train/annotations_without_background.json, img_dir: odinw/PKLot/640/train}
|
||||
train_10_3: {ann_file: odinw/PKLot/640/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_10_30: {ann_file: odinw/PKLot/640/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_10_300: {ann_file: odinw/PKLot/640/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_1_3: {ann_file: odinw/PKLot/640/train/fewshot_train_shot1_seed3.json, img_dir: odinw/PKLot/640/train}
|
||||
train_1_30: {ann_file: odinw/PKLot/640/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_1_300: {ann_file: odinw/PKLot/640/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_3_3: {ann_file: odinw/PKLot/640/train/fewshot_train_shot3_seed3.json, img_dir: odinw/PKLot/640/train}
|
||||
train_3_30: {ann_file: odinw/PKLot/640/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_3_300: {ann_file: odinw/PKLot/640/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_5_3: {ann_file: odinw/PKLot/640/train/fewshot_train_shot5_seed3.json, img_dir: odinw/PKLot/640/train}
|
||||
train_5_30: {ann_file: odinw/PKLot/640/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
train_5_300: {ann_file: odinw/PKLot/640/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/PKLot/640/train}
|
||||
val: {ann_file: odinw/PKLot/640/valid/annotations_without_background.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_10_3: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_10_30: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot10_seed30.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_10_300: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/PKLot/640/valid}
|
||||
val_1_3: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_1_30: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_1_300: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot1_seed300.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_3_3: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_3_30: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_3_300: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot3_seed300.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_5_3: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_5_30: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/PKLot/640/valid}
|
||||
val_5_300: {ann_file: odinw/PKLot/640/valid/fewshot_val_shot5_seed300.json, img_dir: odinw/PKLot/640/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'PKLot_640'
|
||||
|
|
@ -0,0 +1,67 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
CAPTION_PROMPT: '[{"prefix": "there is a ", "name": "package", "suffix": " on the
|
||||
porch"}]'
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "package", "supercategory": "packages"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/Packages/Raw/test/annotations_without_background.json,
|
||||
img_dir: odinw/Packages/Raw/test}
|
||||
train: {ann_file: odinw/Packages/Raw/train/annotations_without_background.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_10_3: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_10_30: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_10_300: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_1_3: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_1_30: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_1_300: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_3_3: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_3_30: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_3_300: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_5_3: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_5_30: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
train_5_300: {ann_file: odinw/Packages/Raw/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/train}
|
||||
val: {ann_file: odinw/Packages/Raw/valid/annotations_without_background.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
val_10_3: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_10_30: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
val_10_300: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
val_1_3: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_1_30: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_1_300: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
val_3_3: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_3_30: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_3_300: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
val_5_3: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_5_30: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/Packages/Raw/valid}
|
||||
val_5_300: {ann_file: odinw/Packages/Raw/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/Packages/Raw/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Packages'
|
||||
|
|
@ -0,0 +1,68 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "aeroplane", "supercategory": "VOC"}, {"id":
|
||||
2, "name": "bicycle", "supercategory": "VOC"}, {"id": 3, "name": "bird", "supercategory":
|
||||
"VOC"}, {"id": 4, "name": "boat", "supercategory": "VOC"}, {"id": 5, "name": "bottle",
|
||||
"supercategory": "VOC"}, {"id": 6, "name": "bus", "supercategory": "VOC"}, {"id":
|
||||
7, "name": "car", "supercategory": "VOC"}, {"id": 8, "name": "cat", "supercategory":
|
||||
"VOC"}, {"id": 9, "name": "chair", "supercategory": "VOC"}, {"id": 10, "name":
|
||||
"cow", "supercategory": "VOC"}, {"id": 11, "name": "diningtable", "supercategory":
|
||||
"VOC"}, {"id": 12, "name": "dog", "supercategory": "VOC"}, {"id": 13, "name":
|
||||
"horse", "supercategory": "VOC"}, {"id": 14, "name": "motorbike", "supercategory":
|
||||
"VOC"}, {"id": 15, "name": "person", "supercategory": "VOC"}, {"id": 16, "name":
|
||||
"pottedplant", "supercategory": "VOC"}, {"id": 17, "name": "sheep", "supercategory":
|
||||
"VOC"}, {"id": 18, "name": "sofa", "supercategory": "VOC"}, {"id": 19, "name":
|
||||
"train", "supercategory": "VOC"}, {"id": 20, "name": "tvmonitor", "supercategory":
|
||||
"VOC"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/PascalVOC/valid/annotations_without_background.json, img_dir: odinw/PascalVOC/valid}
|
||||
train: {ann_file: odinw/PascalVOC/train/annotations_without_background.json, img_dir: odinw/PascalVOC/train}
|
||||
train_10_3: {ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_10_30: {ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_10_300: {ann_file: odinw/PascalVOC/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_1_3: {ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed3.json, img_dir: odinw/PascalVOC/train}
|
||||
train_1_30: {ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_1_300: {ann_file: odinw/PascalVOC/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_3_3: {ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed3.json, img_dir: odinw/PascalVOC/train}
|
||||
train_3_30: {ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_3_300: {ann_file: odinw/PascalVOC/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_5_3: {ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed3.json, img_dir: odinw/PascalVOC/train}
|
||||
train_5_30: {ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
train_5_300: {ann_file: odinw/PascalVOC/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/PascalVOC/train}
|
||||
val: {ann_file: odinw/PascalVOC/valid/annotations_without_background.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_10_3: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_10_30: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed30.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_10_300: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/PascalVOC/valid}
|
||||
val_1_3: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_1_30: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_1_300: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot1_seed300.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_3_3: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_3_30: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_3_300: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot3_seed300.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_5_3: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_5_30: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/PascalVOC/valid}
|
||||
val_5_300: {ann_file: odinw/PascalVOC/valid/fewshot_val_shot5_seed300.json, img_dir: odinw/PascalVOC/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 21}
|
||||
DYHEAD: {NUM_CLASSES: 21}
|
||||
FCOS: {NUM_CLASSES: 21}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 21}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'PascalVOC'
|
||||
|
|
@ -0,0 +1,72 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "raccoon", "supercategory": "raccoons"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/test/annotations_without_background.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/test}
|
||||
train: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/annotations_without_background.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_10_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_10_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_10_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_1_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_1_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_1_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_3_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_3_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_3_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_5_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_5_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
train_5_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/train}
|
||||
val: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/annotations_without_background.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_10_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_10_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_10_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_1_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_1_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_1_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_3_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_3_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_3_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_5_3: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_5_30: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
val_5_300: {ann_file: odinw/Raccoon/Raccoon.v2-raw.coco/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/Raccoon/Raccoon.v2-raw.coco/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'Raccoon_Raccoon'
|
||||
|
|
@ -0,0 +1,74 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Crab", "supercategory": "shellfish"}, {"id":
|
||||
2, "name": "Lobster", "supercategory": "shellfish"}, {"id": 3, "name": "Shrimp",
|
||||
"supercategory": "shellfish"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/ShellfishOpenImages/raw/test/annotations_without_background.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/test}
|
||||
train: {ann_file: odinw/ShellfishOpenImages/raw/train/annotations_without_background.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_10_3: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_10_30: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_10_300: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_1_3: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_1_30: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_1_300: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_3_3: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_3_30: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_3_300: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_5_3: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_5_30: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
train_5_300: {ann_file: odinw/ShellfishOpenImages/raw/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/train}
|
||||
val: {ann_file: odinw/ShellfishOpenImages/raw/valid/annotations_without_background.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_10_3: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_10_30: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_10_300: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_1_3: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_1_30: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_1_300: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_3_3: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_3_30: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_3_300: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_5_3: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_5_30: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
val_5_300: {ann_file: odinw/ShellfishOpenImages/raw/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/ShellfishOpenImages/raw/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 4}
|
||||
DYHEAD: {NUM_CLASSES: 4}
|
||||
FCOS: {NUM_CLASSES: 4}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 4}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'ShellfishOpenImages'
|
||||
|
|
@ -0,0 +1,70 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "cheetah", "supercategory": "cheetah"}, {"id":
|
||||
2, "name": "human", "supercategory": "cheetah"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/ThermalCheetah/test/annotations_without_background.json,
|
||||
img_dir: odinw/ThermalCheetah/test}
|
||||
train: {ann_file: odinw/ThermalCheetah/train/annotations_without_background.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_10_3: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_10_30: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_10_300: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_1_3: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_1_30: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_1_300: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_3_3: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_3_30: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_3_300: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_5_3: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_5_30: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
train_5_300: {ann_file: odinw/ThermalCheetah/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/train}
|
||||
val: {ann_file: odinw/ThermalCheetah/valid/annotations_without_background.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_10_3: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_10_30: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_10_300: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_1_3: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/ThermalCheetah/valid}
|
||||
val_1_30: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_1_300: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_3_3: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/ThermalCheetah/valid}
|
||||
val_3_30: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_3_300: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_5_3: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/ThermalCheetah/valid}
|
||||
val_5_30: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
val_5_300: {ann_file: odinw/ThermalCheetah/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/ThermalCheetah/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'ThermalCheetah'
|
||||
|
|
@ -0,0 +1,101 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "0", "supercategory": "Card-Types"}, {"id":
|
||||
2, "name": "1", "supercategory": "Card-Types"}, {"id": 3, "name": "2", "supercategory":
|
||||
"Card-Types"}, {"id": 4, "name": "3", "supercategory": "Card-Types"}, {"id": 5,
|
||||
"name": "4", "supercategory": "Card-Types"}, {"id": 6, "name": "5", "supercategory":
|
||||
"Card-Types"}, {"id": 7, "name": "6", "supercategory": "Card-Types"}, {"id": 8,
|
||||
"name": "7", "supercategory": "Card-Types"}, {"id": 9, "name": "8", "supercategory":
|
||||
"Card-Types"}, {"id": 10, "name": "9", "supercategory": "Card-Types"}, {"id":
|
||||
11, "name": "10", "supercategory": "Card-Types"}, {"id": 12, "name": "11", "supercategory":
|
||||
"Card-Types"}, {"id": 13, "name": "12", "supercategory": "Card-Types"}, {"id":
|
||||
14, "name": "13", "supercategory": "Card-Types"}, {"id": 15, "name": "14", "supercategory":
|
||||
"Card-Types"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/UnoCards/raw/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_10_3: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_10_30: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_10_300: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_1_3: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_1_30: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_1_300: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_3_3: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_3_30: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_3_300: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_5_3: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_5_30: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
minival_5_300: {ann_file: odinw/UnoCards/raw/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/mini_val}
|
||||
test: {ann_file: odinw/UnoCards/raw/test/annotations_without_background.json,
|
||||
img_dir: odinw/UnoCards/raw/test}
|
||||
train: {ann_file: odinw/UnoCards/raw/train/annotations_without_background.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_10_3: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_10_30: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_10_300: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_1_3: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_1_30: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_1_300: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_3_3: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_3_30: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_3_300: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_5_3: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_5_30: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
train_5_300: {ann_file: odinw/UnoCards/raw/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/train}
|
||||
val: {ann_file: odinw/UnoCards/raw/valid/annotations_without_background.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
val_10_3: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_10_30: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
val_10_300: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
val_1_3: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_1_30: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_1_300: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
val_3_3: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_3_30: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_3_300: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
val_5_3: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_5_30: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/UnoCards/raw/valid}
|
||||
val_5_300: {ann_file: odinw/UnoCards/raw/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/UnoCards/raw/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 16}
|
||||
DYHEAD: {NUM_CLASSES: 16}
|
||||
FCOS: {NUM_CLASSES: 16}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 16}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'UnoCards'
|
||||
|
|
@ -0,0 +1,101 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Ambulance", "supercategory": "vehicles"},
|
||||
{"id": 2, "name": "Bus", "supercategory": "vehicles"}, {"id": 3, "name": "Car",
|
||||
"supercategory": "vehicles"}, {"id": 4, "name": "Motorcycle", "supercategory":
|
||||
"vehicles"}, {"id": 5, "name": "Truck", "supercategory": "vehicles"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_10_3: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_10_30: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_10_300: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_1_3: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_1_30: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_1_300: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_3_3: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_3_30: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_3_300: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_5_3: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_5_30: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
minival_5_300: {ann_file: odinw/VehiclesOpenImages/416x416/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/mini_val}
|
||||
test: {ann_file: odinw/VehiclesOpenImages/416x416/test/annotations_without_background.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/test}
|
||||
train: {ann_file: odinw/VehiclesOpenImages/416x416/train/annotations_without_background.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_10_3: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_10_30: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_10_300: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_1_3: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_1_30: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_1_300: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_3_3: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_3_30: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_3_300: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_5_3: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_5_30: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
train_5_300: {ann_file: odinw/VehiclesOpenImages/416x416/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/train}
|
||||
val: {ann_file: odinw/VehiclesOpenImages/416x416/valid/annotations_without_background.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_10_3: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_10_30: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_10_300: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_1_3: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_1_30: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_1_300: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_3_3: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_3_30: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_3_300: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_5_3: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_5_30: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
val_5_300: {ann_file: odinw/VehiclesOpenImages/416x416/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/VehiclesOpenImages/416x416/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 6}
|
||||
DYHEAD: {NUM_CLASSES: 6}
|
||||
FCOS: {NUM_CLASSES: 6}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 6}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'VehiclesOpenImages'
|
||||
|
|
@ -0,0 +1,69 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "smoke", "supercategory": "Smoke"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/WildfireSmoke/test/annotations_without_background.json,
|
||||
img_dir: odinw/WildfireSmoke/test}
|
||||
train: {ann_file: odinw/WildfireSmoke/train/annotations_without_background.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_10_3: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_10_30: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_10_300: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_1_3: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_1_30: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_1_300: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_3_3: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_3_30: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_3_300: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_5_3: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_5_30: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
train_5_300: {ann_file: odinw/WildfireSmoke/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/train}
|
||||
val: {ann_file: odinw/WildfireSmoke/valid/annotations_without_background.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_10_3: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_10_30: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_10_300: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_1_3: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/WildfireSmoke/valid}
|
||||
val_1_30: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_1_300: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_3_3: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/WildfireSmoke/valid}
|
||||
val_3_30: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_3_300: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_5_3: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/WildfireSmoke/valid}
|
||||
val_5_30: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
val_5_300: {ann_file: odinw/WildfireSmoke/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/WildfireSmoke/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'WildfireSmoke'
|
||||
|
|
@ -0,0 +1 @@
|
|||
["configs/odinw_35/AerialMaritimeDrone_large.yaml","configs/odinw_35/AerialMaritimeDrone_tiled.yaml","configs/odinw_35/AmericanSignLanguageLetters_American_Sign_Language_Letters.v1-v1.coco.yaml","configs/odinw_35/Aquarium_Aquarium_Combined.v2-raw-1024.coco.yaml","configs/odinw_35/BCCD_BCCD.v3-raw.coco.yaml","configs/odinw_35/ChessPieces_Chess_Pieces.v23-raw.coco.yaml","configs/odinw_35/CottontailRabbits.yaml","configs/odinw_35/DroneControl_Drone_Control.v3-raw.coco.yaml","configs/odinw_35/EgoHands_generic.yaml","configs/odinw_35/EgoHands_specific.yaml","configs/odinw_35/HardHatWorkers_raw.yaml","configs/odinw_35/MaskWearing_raw.yaml","configs/odinw_35/MountainDewCommercial.yaml","configs/odinw_35/NorthAmericaMushrooms_North_American_Mushrooms.v1-416x416.coco.yaml","configs/odinw_35/OxfordPets_by-breed.yaml","configs/odinw_35/OxfordPets_by-species.yaml","configs/odinw_35/PKLot_640.yaml","configs/odinw_35/Packages_Raw.yaml","configs/odinw_35/PascalVOC.yaml","configs/odinw_35/Raccoon_Raccoon.v2-raw.coco.yaml","configs/odinw_35/ShellfishOpenImages_raw.yaml","configs/odinw_35/ThermalCheetah.yaml","configs/odinw_35/UnoCards_raw.yaml","configs/odinw_35/VehiclesOpenImages_416x416.yaml","configs/odinw_35/WildfireSmoke.yaml","configs/odinw_35/boggleBoards_416x416AutoOrient_export_.yaml","configs/odinw_35/brackishUnderwater_960x540.yaml","configs/odinw_35/dice_mediumColor_export.yaml","configs/odinw_35/openPoetryVision_512x512.yaml","configs/odinw_35/pistols_export.yaml","configs/odinw_35/plantdoc_416x416.yaml","configs/odinw_35/pothole.yaml","configs/odinw_35/selfdrivingCar_fixedLarge_export_.yaml","configs/odinw_35/thermalDogsAndPeople.yaml","configs/odinw_35/websiteScreenshots.yaml"]
|
||||
|
|
@ -0,0 +1,94 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Q", "supercategory": "letters"}, {"id":
|
||||
2, "name": "a", "supercategory": "letters"}, {"id": 3, "name": "an", "supercategory":
|
||||
"letters"}, {"id": 4, "name": "b", "supercategory": "letters"}, {"id": 5, "name":
|
||||
"c", "supercategory": "letters"}, {"id": 6, "name": "d", "supercategory": "letters"},
|
||||
{"id": 7, "name": "e", "supercategory": "letters"}, {"id": 8, "name": "er", "supercategory":
|
||||
"letters"}, {"id": 9, "name": "f", "supercategory": "letters"}, {"id": 10, "name":
|
||||
"g", "supercategory": "letters"}, {"id": 11, "name": "h", "supercategory": "letters"},
|
||||
{"id": 12, "name": "he", "supercategory": "letters"}, {"id": 13, "name": "i",
|
||||
"supercategory": "letters"}, {"id": 14, "name": "in", "supercategory": "letters"},
|
||||
{"id": 15, "name": "j", "supercategory": "letters"}, {"id": 16, "name": "k", "supercategory":
|
||||
"letters"}, {"id": 17, "name": "l", "supercategory": "letters"}, {"id": 18, "name":
|
||||
"m", "supercategory": "letters"}, {"id": 19, "name": "n", "supercategory": "letters"},
|
||||
{"id": 20, "name": "o", "supercategory": "letters"}, {"id": 21, "name": "o ",
|
||||
"supercategory": "letters"}, {"id": 22, "name": "p", "supercategory": "letters"},
|
||||
{"id": 23, "name": "q", "supercategory": "letters"}, {"id": 24, "name": "qu",
|
||||
"supercategory": "letters"}, {"id": 25, "name": "r", "supercategory": "letters"},
|
||||
{"id": 26, "name": "s", "supercategory": "letters"}, {"id": 27, "name": "t", "supercategory":
|
||||
"letters"}, {"id": 28, "name": "t\\", "supercategory": "letters"}, {"id": 29,
|
||||
"name": "th", "supercategory": "letters"}, {"id": 30, "name": "u", "supercategory":
|
||||
"letters"}, {"id": 31, "name": "v", "supercategory": "letters"}, {"id": 32, "name":
|
||||
"w", "supercategory": "letters"}, {"id": 33, "name": "wild", "supercategory":
|
||||
"letters"}, {"id": 34, "name": "x", "supercategory": "letters"}, {"id": 35, "name":
|
||||
"y", "supercategory": "letters"}, {"id": 36, "name": "z", "supercategory": "letters"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/test_annotations_without_background.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/train_annotations_without_background.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_10_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_10_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_10_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_1_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_1_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_1_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_3_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_3_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_3_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_5_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_5_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
train_5_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/val_annotations_without_background.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_10_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_10_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_10_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_1_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_1_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_1_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_3_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_3_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_3_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_5_3: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_5_30: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
val_5_300: {ann_file: odinw/boggleBoards/416x416AutoOrient/export/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/boggleBoards/416x416AutoOrient/export/}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 35}
|
||||
DYHEAD: {NUM_CLASSES: 35}
|
||||
FCOS: {NUM_CLASSES: 35}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 35}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'boggleBoards'
|
||||
|
|
@ -0,0 +1,104 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "crab", "supercategory": "animals"}, {"id":
|
||||
2, "name": "fish", "supercategory": "animals"}, {"id": 3, "name": "jellyfish",
|
||||
"supercategory": "animals"}, {"id": 4, "name": "shrimp", "supercategory": "animals"},
|
||||
{"id": 5, "name": "small_fish", "supercategory": "animals"}, {"id": 6, "name":
|
||||
"starfish", "supercategory": "animals"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/brackishUnderwater/960x540/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_10_3: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_10_30: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_10_300: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_1_3: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_1_30: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_1_300: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_3_3: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_3_30: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_3_300: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_3shot_3seed: {ann_file: odinw/original/brackishUnderwater/960x540/mini_val/annotations_without_background_3shot_3seed.json,
|
||||
img_dir: odinw/original/brackishUnderwater/960x540/mini_val}
|
||||
minival_5_3: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_5_30: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
minival_5_300: {ann_file: odinw/brackishUnderwater/960x540/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/mini_val}
|
||||
test: {ann_file: odinw/brackishUnderwater/960x540/test/annotations_without_background.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/test}
|
||||
train: {ann_file: odinw/brackishUnderwater/960x540/train/annotations_without_background.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_10_3: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_10_30: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_10_300: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_1_3: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_1_30: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_1_300: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_3_3: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_3_30: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_3_300: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_5_3: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_5_30: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
train_5_300: {ann_file: odinw/brackishUnderwater/960x540/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/train}
|
||||
val: {ann_file: odinw/brackishUnderwater/960x540/valid/annotations_without_background.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_10_3: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_10_30: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_10_300: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_1_3: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_1_30: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_1_300: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_3_3: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_3_30: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_3_300: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_5_3: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_5_30: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
val_5_300: {ann_file: odinw/brackishUnderwater/960x540/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/brackishUnderwater/960x540/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 7}
|
||||
DYHEAD: {NUM_CLASSES: 7}
|
||||
FCOS: {NUM_CLASSES: 7}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 7}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'brackishUnderwater'
|
||||
|
|
@ -0,0 +1,75 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "1", "supercategory": "dice"}, {"id": 2,
|
||||
"name": "2", "supercategory": "dice"}, {"id": 3, "name": "3", "supercategory":
|
||||
"dice"}, {"id": 4, "name": "4", "supercategory": "dice"}, {"id": 5, "name": "5",
|
||||
"supercategory": "dice"}, {"id": 6, "name": "6", "supercategory": "dice"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/dice/mediumColor/export/test_annotations_without_background.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train: {ann_file: odinw/dice/mediumColor/export/train_annotations_without_background.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_10_3: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_10_30: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_10_300: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_1_3: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_1_30: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_1_300: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_3_3: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_3_30: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_3_300: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_5_3: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_5_30: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
train_5_300: {ann_file: odinw/dice/mediumColor/export/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val: {ann_file: odinw/dice/mediumColor/export/val_annotations_without_background.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_10_3: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_10_30: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_10_300: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_1_3: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_1_30: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_1_300: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_3_3: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_3_30: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_3_300: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_5_3: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_5_30: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
val_5_300: {ann_file: odinw/dice/mediumColor/export/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/dice/mediumColor/export}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 71}
|
||||
DYHEAD: {NUM_CLASSES: 71}
|
||||
FCOS: {NUM_CLASSES: 71}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 71}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'dice_mediumColor'
|
||||
|
|
@ -0,0 +1,128 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 2
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "American Typewriter", "supercategory": "text"},
|
||||
{"id": 2, "name": "Andale Mono", "supercategory": "text"}, {"id": 3, "name": "Apple
|
||||
Chancery", "supercategory": "text"}, {"id": 4, "name": "Arial", "supercategory":
|
||||
"text"}, {"id": 5, "name": "Avenir", "supercategory": "text"}, {"id": 6, "name":
|
||||
"Baskerville", "supercategory": "text"}, {"id": 7, "name": "Big Caslon", "supercategory":
|
||||
"text"}, {"id": 8, "name": "Bradley Hand", "supercategory": "text"}, {"id": 9,
|
||||
"name": "Brush Script MT", "supercategory": "text"}, {"id": 10, "name": "Chalkboard",
|
||||
"supercategory": "text"}, {"id": 11, "name": "Comic Sans MS", "supercategory":
|
||||
"text"}, {"id": 12, "name": "Copperplate", "supercategory": "text"}, {"id": 13,
|
||||
"name": "Courier", "supercategory": "text"}, {"id": 14, "name": "Didot", "supercategory":
|
||||
"text"}, {"id": 15, "name": "Futura", "supercategory": "text"}, {"id": 16, "name":
|
||||
"Geneva", "supercategory": "text"}, {"id": 17, "name": "Georgia", "supercategory":
|
||||
"text"}, {"id": 18, "name": "Gill Sans", "supercategory": "text"}, {"id": 19,
|
||||
"name": "Helvetica", "supercategory": "text"}, {"id": 20, "name": "Herculanum",
|
||||
"supercategory": "text"}, {"id": 21, "name": "Impact", "supercategory": "text"},
|
||||
{"id": 22, "name": "Kefa", "supercategory": "text"}, {"id": 23, "name": "Lucida
|
||||
Grande", "supercategory": "text"}, {"id": 24, "name": "Luminari", "supercategory":
|
||||
"text"}, {"id": 25, "name": "Marker Felt", "supercategory": "text"}, {"id": 26,
|
||||
"name": "Menlo", "supercategory": "text"}, {"id": 27, "name": "Monaco", "supercategory":
|
||||
"text"}, {"id": 28, "name": "Noteworthy", "supercategory": "text"}, {"id": 29,
|
||||
"name": "Optima", "supercategory": "text"}, {"id": 30, "name": "PT Sans", "supercategory":
|
||||
"text"}, {"id": 31, "name": "PT Serif", "supercategory": "text"}, {"id": 32, "name":
|
||||
"Palatino", "supercategory": "text"}, {"id": 33, "name": "Papyrus", "supercategory":
|
||||
"text"}, {"id": 34, "name": "Phosphate", "supercategory": "text"}, {"id": 35,
|
||||
"name": "Rockwell", "supercategory": "text"}, {"id": 36, "name": "SF Pro", "supercategory":
|
||||
"text"}, {"id": 37, "name": "SignPainter", "supercategory": "text"}, {"id": 38,
|
||||
"name": "Skia", "supercategory": "text"}, {"id": 39, "name": "Snell Roundhand",
|
||||
"supercategory": "text"}, {"id": 40, "name": "Tahoma", "supercategory": "text"},
|
||||
{"id": 41, "name": "Times New Roman", "supercategory": "text"}, {"id": 42, "name":
|
||||
"Trebuchet MS", "supercategory": "text"}, {"id": 43, "name": "Verdana", "supercategory":
|
||||
"text"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/openPoetryVision/512x512/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_10_3: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_10_30: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_10_300: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_1_3: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_1_30: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_1_300: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_3_3: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_3_30: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_3_300: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_5_3: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_5_30: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
minival_5_300: {ann_file: odinw/openPoetryVision/512x512/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/mini_val}
|
||||
test: {ann_file: odinw/openPoetryVision/512x512/test/annotations_without_background.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/test}
|
||||
train: {ann_file: odinw/openPoetryVision/512x512/train/annotations_without_background.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_10_3: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_10_30: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_10_300: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_1_3: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_1_30: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_1_300: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_3_3: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_3_30: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_3_300: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_5_3: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_5_30: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
train_5_300: {ann_file: odinw/openPoetryVision/512x512/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/train}
|
||||
val: {ann_file: odinw/openPoetryVision/512x512/valid/annotations_without_background.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_10_3: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_10_30: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_10_300: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_1_3: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_1_30: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_1_300: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_3_3: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_3_30: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_3_300: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_5_3: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_5_30: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
val_5_300: {ann_file: odinw/openPoetryVision/512x512/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/openPoetryVision/512x512/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 44}
|
||||
DYHEAD: {NUM_CLASSES: 44}
|
||||
FCOS: {NUM_CLASSES: 44}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 44}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'openPoetryVision'
|
||||
|
|
@ -0,0 +1,53 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "pistol", "supercategory": "Guns"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/pistols/export/test_annotations_without_background.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train: {ann_file: odinw/pistols/export/train_annotations_without_background.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train_10_3: {ann_file: odinw/pistols/export/fewshot_train_shot10_seed3.json, img_dir: odinw/pistols/export}
|
||||
train_10_30: {ann_file: odinw/pistols/export/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train_10_300: {ann_file: odinw/pistols/export/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train_1_3: {ann_file: odinw/pistols/export/fewshot_train_shot1_seed3.json, img_dir: odinw/pistols/export}
|
||||
train_1_30: {ann_file: odinw/pistols/export/fewshot_train_shot1_seed30.json, img_dir: odinw/pistols/export}
|
||||
train_1_300: {ann_file: odinw/pistols/export/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train_3_3: {ann_file: odinw/pistols/export/fewshot_train_shot3_seed3.json, img_dir: odinw/pistols/export}
|
||||
train_3_30: {ann_file: odinw/pistols/export/fewshot_train_shot3_seed30.json, img_dir: odinw/pistols/export}
|
||||
train_3_300: {ann_file: odinw/pistols/export/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
train_5_3: {ann_file: odinw/pistols/export/fewshot_train_shot5_seed3.json, img_dir: odinw/pistols/export}
|
||||
train_5_30: {ann_file: odinw/pistols/export/fewshot_train_shot5_seed30.json, img_dir: odinw/pistols/export}
|
||||
train_5_300: {ann_file: odinw/pistols/export/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
val: {ann_file: odinw/pistols/export/val_annotations_without_background.json,
|
||||
img_dir: odinw/pistols/export}
|
||||
val_10_3: {ann_file: odinw/pistols/export/fewshot_val_shot10_seed3.json, img_dir: odinw/pistols/export}
|
||||
val_10_30: {ann_file: odinw/pistols/export/fewshot_val_shot10_seed30.json, img_dir: odinw/pistols/export}
|
||||
val_10_300: {ann_file: odinw/pistols/export/fewshot_val_shot10_seed300.json, img_dir: odinw/pistols/export}
|
||||
val_1_3: {ann_file: odinw/pistols/export/fewshot_val_shot1_seed3.json, img_dir: odinw/pistols/export}
|
||||
val_1_30: {ann_file: odinw/pistols/export/fewshot_val_shot1_seed30.json, img_dir: odinw/pistols/export}
|
||||
val_1_300: {ann_file: odinw/pistols/export/fewshot_val_shot1_seed300.json, img_dir: odinw/pistols/export}
|
||||
val_3_3: {ann_file: odinw/pistols/export/fewshot_val_shot3_seed3.json, img_dir: odinw/pistols/export}
|
||||
val_3_30: {ann_file: odinw/pistols/export/fewshot_val_shot3_seed30.json, img_dir: odinw/pistols/export}
|
||||
val_3_300: {ann_file: odinw/pistols/export/fewshot_val_shot3_seed300.json, img_dir: odinw/pistols/export}
|
||||
val_5_3: {ann_file: odinw/pistols/export/fewshot_val_shot5_seed3.json, img_dir: odinw/pistols/export}
|
||||
val_5_30: {ann_file: odinw/pistols/export/fewshot_val_shot5_seed30.json, img_dir: odinw/pistols/export}
|
||||
val_5_300: {ann_file: odinw/pistols/export/fewshot_val_shot5_seed300.json, img_dir: odinw/pistols/export}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 297}
|
||||
DYHEAD: {NUM_CLASSES: 297}
|
||||
FCOS: {NUM_CLASSES: 297}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 297}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'pistols'
|
||||
|
|
@ -0,0 +1,136 @@
|
|||
DATALOADER:
|
||||
ASPECT_RATIO_GROUPING: false
|
||||
SIZE_DIVISIBILITY: 32
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Apple Scab Leaf", "supercategory": "leaves"},
|
||||
{"id": 2, "name": "Apple leaf", "supercategory": "leaves"}, {"id": 3, "name":
|
||||
"Apple rust leaf", "supercategory": "leaves"}, {"id": 4, "name": "Bell_pepper
|
||||
leaf", "supercategory": "leaves"}, {"id": 5, "name": "Bell_pepper leaf spot",
|
||||
"supercategory": "leaves"}, {"id": 6, "name": "Blueberry leaf", "supercategory":
|
||||
"leaves"}, {"id": 7, "name": "Cherry leaf", "supercategory": "leaves"}, {"id":
|
||||
8, "name": "Corn Gray leaf spot", "supercategory": "leaves"}, {"id": 9, "name":
|
||||
"Corn leaf blight", "supercategory": "leaves"}, {"id": 10, "name": "Corn rust
|
||||
leaf", "supercategory": "leaves"}, {"id": 11, "name": "Peach leaf", "supercategory":
|
||||
"leaves"}, {"id": 12, "name": "Potato leaf", "supercategory": "leaves"}, {"id":
|
||||
13, "name": "Potato leaf early blight", "supercategory": "leaves"}, {"id": 14,
|
||||
"name": "Potato leaf late blight", "supercategory": "leaves"}, {"id": 15, "name":
|
||||
"Raspberry leaf", "supercategory": "leaves"}, {"id": 16, "name": "Soyabean leaf",
|
||||
"supercategory": "leaves"}, {"id": 17, "name": "Soybean leaf", "supercategory":
|
||||
"leaves"}, {"id": 18, "name": "Squash Powdery mildew leaf", "supercategory": "leaves"},
|
||||
{"id": 19, "name": "Strawberry leaf", "supercategory": "leaves"}, {"id": 20, "name":
|
||||
"Tomato Early blight leaf", "supercategory": "leaves"}, {"id": 21, "name": "Tomato
|
||||
Septoria leaf spot", "supercategory": "leaves"}, {"id": 22, "name": "Tomato leaf",
|
||||
"supercategory": "leaves"}, {"id": 23, "name": "Tomato leaf bacterial spot", "supercategory":
|
||||
"leaves"}, {"id": 24, "name": "Tomato leaf late blight", "supercategory": "leaves"},
|
||||
{"id": 25, "name": "Tomato leaf mosaic virus", "supercategory": "leaves"}, {"id":
|
||||
26, "name": "Tomato leaf yellow virus", "supercategory": "leaves"}, {"id": 27,
|
||||
"name": "Tomato mold leaf", "supercategory": "leaves"}, {"id": 28, "name": "Tomato
|
||||
two spotted spider mites leaf", "supercategory": "leaves"}, {"id": 29, "name":
|
||||
"grape leaf", "supercategory": "leaves"}, {"id": 30, "name": "grape leaf black
|
||||
rot", "supercategory": "leaves"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test:
|
||||
ann_file: odinw/plantdoc/100x100/test/annotations_without_background.json
|
||||
img_dir: odinw/plantdoc/100x100/test
|
||||
train:
|
||||
ann_file: odinw/plantdoc/100x100/train/annotations_without_background.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_10_3:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot10_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_10_30:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot10_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_10_300:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot10_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_1_3:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot1_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_1_30:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot1_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_1_300:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot1_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_3_3:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot3_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_3_30:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot3_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_3_300:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot3_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_5_3:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot5_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_5_30:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot5_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
train_5_300:
|
||||
ann_file: odinw/plantdoc/100x100/train/fewshot_train_shot5_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/train
|
||||
val:
|
||||
ann_file: odinw/plantdoc/100x100/valid/annotations_without_background.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_10_3:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot10_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_10_30:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot10_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_10_300:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot10_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_1_3:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot1_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_1_30:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot1_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_1_300:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot1_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_3_3:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot3_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_3_30:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot3_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_3_300:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot3_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_5_3:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot5_seed3.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_5_30:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot5_seed30.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
val_5_300:
|
||||
ann_file: odinw/plantdoc/100x100/valid/fewshot_val_shot5_seed300.json
|
||||
img_dir: odinw/plantdoc/100x100/valid
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT:
|
||||
MAX_SIZE_TEST: 1333
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MODEL:
|
||||
ATSS:
|
||||
NUM_CLASSES: 31
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 31
|
||||
FCOS:
|
||||
NUM_CLASSES: 31
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 31
|
||||
SOLVER:
|
||||
CHECKPOINT_PERIOD: 100
|
||||
MAX_EPOCH: 12
|
||||
WARMUP_ITERS: 0
|
||||
TEST:
|
||||
IMS_PER_BATCH: 8
|
||||
|
|
@ -0,0 +1,97 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 4
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "Apple Scab Leaf", "supercategory": "leaves"},
|
||||
{"id": 2, "name": "Apple leaf", "supercategory": "leaves"}, {"id": 3, "name":
|
||||
"Apple rust leaf", "supercategory": "leaves"}, {"id": 4, "name": "Bell_pepper
|
||||
leaf", "supercategory": "leaves"}, {"id": 5, "name": "Bell_pepper leaf spot",
|
||||
"supercategory": "leaves"}, {"id": 6, "name": "Blueberry leaf", "supercategory":
|
||||
"leaves"}, {"id": 7, "name": "Cherry leaf", "supercategory": "leaves"}, {"id":
|
||||
8, "name": "Corn Gray leaf spot", "supercategory": "leaves"}, {"id": 9, "name":
|
||||
"Corn leaf blight", "supercategory": "leaves"}, {"id": 10, "name": "Corn rust
|
||||
leaf", "supercategory": "leaves"}, {"id": 11, "name": "Peach leaf", "supercategory":
|
||||
"leaves"}, {"id": 12, "name": "Potato leaf", "supercategory": "leaves"}, {"id":
|
||||
13, "name": "Potato leaf early blight", "supercategory": "leaves"}, {"id": 14,
|
||||
"name": "Potato leaf late blight", "supercategory": "leaves"}, {"id": 15, "name":
|
||||
"Raspberry leaf", "supercategory": "leaves"}, {"id": 16, "name": "Soyabean leaf",
|
||||
"supercategory": "leaves"}, {"id": 17, "name": "Soybean leaf", "supercategory":
|
||||
"leaves"}, {"id": 18, "name": "Squash Powdery mildew leaf", "supercategory": "leaves"},
|
||||
{"id": 19, "name": "Strawberry leaf", "supercategory": "leaves"}, {"id": 20, "name":
|
||||
"Tomato Early blight leaf", "supercategory": "leaves"}, {"id": 21, "name": "Tomato
|
||||
Septoria leaf spot", "supercategory": "leaves"}, {"id": 22, "name": "Tomato leaf",
|
||||
"supercategory": "leaves"}, {"id": 23, "name": "Tomato leaf bacterial spot", "supercategory":
|
||||
"leaves"}, {"id": 24, "name": "Tomato leaf late blight", "supercategory": "leaves"},
|
||||
{"id": 25, "name": "Tomato leaf mosaic virus", "supercategory": "leaves"}, {"id":
|
||||
26, "name": "Tomato leaf yellow virus", "supercategory": "leaves"}, {"id": 27,
|
||||
"name": "Tomato mold leaf", "supercategory": "leaves"}, {"id": 28, "name": "Tomato
|
||||
two spotted spider mites leaf", "supercategory": "leaves"}, {"id": 29, "name":
|
||||
"grape leaf", "supercategory": "leaves"}, {"id": 30, "name": "grape leaf black
|
||||
rot", "supercategory": "leaves"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/plantdoc/416x416/test/annotations_without_background.json,
|
||||
img_dir: odinw/plantdoc/416x416/test}
|
||||
train: {ann_file: odinw/plantdoc/416x416/train/annotations_without_background.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_10_3: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_10_30: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_10_300: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_1_3: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_1_30: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_1_300: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_3_3: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_3_30: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_3_300: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_5_3: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_5_30: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
train_5_300: {ann_file: odinw/plantdoc/416x416/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/train}
|
||||
val: {ann_file: odinw/plantdoc/416x416/valid/annotations_without_background.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_10_3: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_10_30: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_10_300: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_1_3: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_1_30: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_1_300: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_3_3: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_3_30: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_3_300: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_5_3: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_5_30: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
val_5_300: {ann_file: odinw/plantdoc/416x416/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/plantdoc/416x416/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 31}
|
||||
DYHEAD: {NUM_CLASSES: 31}
|
||||
FCOS: {NUM_CLASSES: 31}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 31}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'plantdoc'
|
||||
|
|
@ -0,0 +1,52 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
CAPTION_PROMPT: '[{"prefix": "there are some ", "name": "holes", "suffix": " on
|
||||
the road"}]'
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "pothole", "supercategory": "potholes"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/pothole/test/annotations_without_background.json, img_dir: odinw/pothole/test}
|
||||
train: {ann_file: odinw/pothole/train/annotations_without_background.json, img_dir: odinw/pothole/train}
|
||||
train_10_3: {ann_file: odinw/pothole/train/fewshot_train_shot10_seed3.json, img_dir: odinw/pothole/train}
|
||||
train_10_30: {ann_file: odinw/pothole/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/pothole/train}
|
||||
train_10_300: {ann_file: odinw/pothole/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/pothole/train}
|
||||
train_1_3: {ann_file: odinw/pothole/train/fewshot_train_shot1_seed3.json, img_dir: odinw/pothole/train}
|
||||
train_1_30: {ann_file: odinw/pothole/train/fewshot_train_shot1_seed30.json, img_dir: odinw/pothole/train}
|
||||
train_1_300: {ann_file: odinw/pothole/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/pothole/train}
|
||||
train_3_3: {ann_file: odinw/pothole/train/fewshot_train_shot3_seed3.json, img_dir: odinw/pothole/train}
|
||||
train_3_30: {ann_file: odinw/pothole/train/fewshot_train_shot3_seed30.json, img_dir: odinw/pothole/train}
|
||||
train_3_300: {ann_file: odinw/pothole/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/pothole/train}
|
||||
train_5_3: {ann_file: odinw/pothole/train/fewshot_train_shot5_seed3.json, img_dir: odinw/pothole/train}
|
||||
train_5_30: {ann_file: odinw/pothole/train/fewshot_train_shot5_seed30.json, img_dir: odinw/pothole/train}
|
||||
train_5_300: {ann_file: odinw/pothole/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/pothole/train}
|
||||
val: {ann_file: odinw/pothole/valid/annotations_without_background.json, img_dir: odinw/pothole/valid}
|
||||
val_10_3: {ann_file: odinw/pothole/valid/fewshot_val_shot10_seed3.json, img_dir: odinw/pothole/valid}
|
||||
val_10_30: {ann_file: odinw/pothole/valid/fewshot_val_shot10_seed30.json, img_dir: odinw/pothole/valid}
|
||||
val_10_300: {ann_file: odinw/pothole/valid/fewshot_val_shot10_seed300.json, img_dir: odinw/pothole/valid}
|
||||
val_1_3: {ann_file: odinw/pothole/valid/fewshot_val_shot1_seed3.json, img_dir: odinw/pothole/valid}
|
||||
val_1_30: {ann_file: odinw/pothole/valid/fewshot_val_shot1_seed30.json, img_dir: odinw/pothole/valid}
|
||||
val_1_300: {ann_file: odinw/pothole/valid/fewshot_val_shot1_seed300.json, img_dir: odinw/pothole/valid}
|
||||
val_3_3: {ann_file: odinw/pothole/valid/fewshot_val_shot3_seed3.json, img_dir: odinw/pothole/valid}
|
||||
val_3_30: {ann_file: odinw/pothole/valid/fewshot_val_shot3_seed30.json, img_dir: odinw/pothole/valid}
|
||||
val_3_300: {ann_file: odinw/pothole/valid/fewshot_val_shot3_seed300.json, img_dir: odinw/pothole/valid}
|
||||
val_5_3: {ann_file: odinw/pothole/valid/fewshot_val_shot5_seed3.json, img_dir: odinw/pothole/valid}
|
||||
val_5_30: {ann_file: odinw/pothole/valid/fewshot_val_shot5_seed30.json, img_dir: odinw/pothole/valid}
|
||||
val_5_300: {ann_file: odinw/pothole/valid/fewshot_val_shot5_seed300.json, img_dir: odinw/pothole/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 2}
|
||||
DYHEAD: {NUM_CLASSES: 2}
|
||||
FCOS: {NUM_CLASSES: 2}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 2}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'pothole'
|
||||
|
|
@ -0,0 +1,80 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "biker", "supercategory": "obstacles"}, {"id":
|
||||
2, "name": "car", "supercategory": "obstacles"}, {"id": 3, "name": "pedestrian",
|
||||
"supercategory": "obstacles"}, {"id": 4, "name": "trafficLight", "supercategory":
|
||||
"obstacles"}, {"id": 5, "name": "trafficLight-Green", "supercategory": "obstacles"},
|
||||
{"id": 6, "name": "trafficLight-GreenLeft", "supercategory": "obstacles"}, {"id":
|
||||
7, "name": "trafficLight-Red", "supercategory": "obstacles"}, {"id": 8, "name":
|
||||
"trafficLight-RedLeft", "supercategory": "obstacles"}, {"id": 9, "name": "trafficLight-Yellow",
|
||||
"supercategory": "obstacles"}, {"id": 10, "name": "trafficLight-YellowLeft", "supercategory":
|
||||
"obstacles"}, {"id": 11, "name": "truck", "supercategory": "obstacles"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/selfdrivingCar/fixedLarge/export/test_annotations_without_background.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train: {ann_file: odinw/selfdrivingCar/fixedLarge/export/train_annotations_without_background.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_10_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_10_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_10_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_1_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_1_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_1_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_3_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_3_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_3_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_5_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_5_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
train_5_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val: {ann_file: odinw/selfdrivingCar/fixedLarge/export/val_annotations_without_background.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_10_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_10_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_10_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_1_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_1_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_1_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_3_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_3_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_3_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_5_3: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_5_30: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
val_5_300: {ann_file: odinw/selfdrivingCar/fixedLarge/export/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/selfdrivingCar/fixedLarge/export/}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3000}
|
||||
DYHEAD: {NUM_CLASSES: 3000}
|
||||
FCOS: {NUM_CLASSES: 3000}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3000}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'selfdrivingCar'
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "dog", "supercategory": "dogs-person"}, {"id":
|
||||
2, "name": "person", "supercategory": "dogs-person"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
test: {ann_file: odinw/thermalDogsAndPeople/test/annotations_without_background.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/test}
|
||||
train: {ann_file: odinw/thermalDogsAndPeople/train/annotations_without_background.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_10_3: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_10_30: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_10_300: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_1_3: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_1_30: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_1_300: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_3_3: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_3_30: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_3_300: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_5_3: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_5_30: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
train_5_300: {ann_file: odinw/thermalDogsAndPeople/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/train}
|
||||
val: {ann_file: odinw/thermalDogsAndPeople/valid/annotations_without_background.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_10_3: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_10_30: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_10_300: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_1_3: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_1_30: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_1_300: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_3_3: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_3_30: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_3_300: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_5_3: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_5_30: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
val_5_300: {ann_file: odinw/thermalDogsAndPeople/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/thermalDogsAndPeople/valid}
|
||||
TEST: ("val",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 3}
|
||||
DYHEAD: {NUM_CLASSES: 3}
|
||||
FCOS: {NUM_CLASSES: 3}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 3}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'thermalDogsAndPeople'
|
||||
|
|
@ -0,0 +1,103 @@
|
|||
DATALOADER: {ASPECT_RATIO_GROUPING: false, SIZE_DIVISIBILITY: 32}
|
||||
DATASETS:
|
||||
GENERAL_COPY: 16
|
||||
OVERRIDE_CATEGORY: '[{"id": 1, "name": "button", "supercategory": "elements"}, {"id":
|
||||
2, "name": "field", "supercategory": "elements"}, {"id": 3, "name": "heading",
|
||||
"supercategory": "elements"}, {"id": 4, "name": "iframe", "supercategory": "elements"},
|
||||
{"id": 5, "name": "image", "supercategory": "elements"}, {"id": 6, "name": "label",
|
||||
"supercategory": "elements"}, {"id": 7, "name": "link", "supercategory": "elements"},
|
||||
{"id": 8, "name": "text", "supercategory": "elements"}]'
|
||||
PREDEFINED_TEXT: odinw/pothole/category_description.json
|
||||
REGISTER:
|
||||
minival: {ann_file: odinw/websiteScreenshots/mini_val/annotations_without_background.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_10_3: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot10_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_10_30: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot10_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_10_300: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot10_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_1_3: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot1_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_1_30: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot1_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_1_300: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot1_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_3_3: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot3_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_3_30: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot3_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_3_300: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot3_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_5_3: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot5_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_5_30: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot5_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
minival_5_300: {ann_file: odinw/websiteScreenshots/mini_val/fewshot_minival_shot5_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/mini_val}
|
||||
test: {ann_file: odinw/websiteScreenshots/test/annotations_without_background.json,
|
||||
img_dir: odinw/websiteScreenshots/test}
|
||||
train: {ann_file: odinw/websiteScreenshots/train/annotations_without_background.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_10_3: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot10_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_10_30: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot10_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_10_300: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot10_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_1_3: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot1_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_1_30: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot1_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_1_300: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot1_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_3_3: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot3_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_3_30: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot3_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_3_300: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot3_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_5_3: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot5_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_5_30: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot5_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
train_5_300: {ann_file: odinw/websiteScreenshots/train/fewshot_train_shot5_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/train}
|
||||
val: {ann_file: odinw/websiteScreenshots/valid/annotations_without_background.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_10_3: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot10_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_10_30: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot10_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_10_300: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot10_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_1_3: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot1_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_1_30: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot1_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_1_300: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot1_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_3_3: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot3_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_3_30: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot3_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_3_300: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot3_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_5_3: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot5_seed3.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_5_30: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot5_seed30.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
val_5_300: {ann_file: odinw/websiteScreenshots/valid/fewshot_val_shot5_seed300.json,
|
||||
img_dir: odinw/websiteScreenshots/valid}
|
||||
TEST: ("minival",)
|
||||
TRAIN: ("train",)
|
||||
INPUT: {MAX_SIZE_TEST: 1333, MAX_SIZE_TRAIN: 1333, MIN_SIZE_TEST: 800, MIN_SIZE_TRAIN: 800}
|
||||
MODEL:
|
||||
ATSS: {NUM_CLASSES: 9}
|
||||
DYHEAD: {NUM_CLASSES: 9}
|
||||
FCOS: {NUM_CLASSES: 9}
|
||||
ROI_BOX_HEAD: {NUM_CLASSES: 9}
|
||||
SOLVER: {CHECKPOINT_PERIOD: 100, MAX_EPOCH: 12, WARMUP_ITERS: 0}
|
||||
TEST: {IMS_PER_BATCH: 8}
|
||||
VISION_QUERY:
|
||||
DATASET_NAME: 'websiteScreenshots'
|
||||
|
|
@ -0,0 +1,164 @@
|
|||
MODEL:
|
||||
META_ARCHITECTURE: "GeneralizedVLRCNN_New"
|
||||
WEIGHT: "MODEL/glip_large_model.pth"
|
||||
RPN_ONLY: True
|
||||
RPN_ARCHITECTURE: "VLDYHEAD"
|
||||
|
||||
BACKBONE:
|
||||
CONV_BODY: "SWINT-FPN-RETINANET"
|
||||
OUT_CHANNELS: 256
|
||||
|
||||
SWINT:
|
||||
EMBED_DIM: 192
|
||||
DEPTHS: (2, 2, 18, 2)
|
||||
NUM_HEADS: (6, 12, 24, 48)
|
||||
WINDOW_SIZE: 12
|
||||
OUT_CHANNELS: (192, 384, 768, 1536)
|
||||
DROP_PATH_RATE: 0.4
|
||||
|
||||
LANGUAGE_BACKBONE:
|
||||
FREEZE: False
|
||||
TOKENIZER_TYPE: "bert-base-uncased"
|
||||
MODEL_TYPE: "bert-base-uncased" # "roberta-base", "clip"
|
||||
# TOKENIZER_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased"
|
||||
# MODEL_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased" # "roberta-base", "clip"
|
||||
MASK_SPECIAL: False
|
||||
|
||||
ROI_BOX_HEAD:
|
||||
POOLER_RESOLUTION: 7
|
||||
POOLER_SCALES: (0.125, 0.0625, 0.03125, 0.015625, 0.0078125) # TODO: check
|
||||
POOLER_SAMPLING_RATIO: 0
|
||||
|
||||
RPN:
|
||||
USE_FPN: True
|
||||
ANCHOR_SIZES: (64, 128, 256, 512, 1024)
|
||||
ANCHOR_STRIDE: (8, 16, 32, 64, 128)
|
||||
ASPECT_RATIOS: (1.0,)
|
||||
SCALES_PER_OCTAVE: 1
|
||||
|
||||
DYHEAD:
|
||||
CHANNELS: 256
|
||||
NUM_CONVS: 8
|
||||
USE_GN: True
|
||||
USE_DYRELU: True
|
||||
USE_DFCONV: True
|
||||
USE_DYFUSE: True
|
||||
TOPK: 9 # topk for selecting candidate positive samples from each level
|
||||
SCORE_AGG: "MEAN"
|
||||
LOG_SCALE: 0.0
|
||||
|
||||
# USE_CHECKPOINT: True
|
||||
USE_CHECKPOINT: False
|
||||
FUSE_CONFIG:
|
||||
USE_FUSED_FEATURES_DOT_PRODUCT: True
|
||||
EARLY_FUSE_ON: True
|
||||
TYPE: "MHA-B"
|
||||
USE_CLASSIFICATION_LOSS: False
|
||||
USE_TOKEN_LOSS: False
|
||||
USE_CONTRASTIVE_ALIGN_LOSS: False
|
||||
CONTRASTIVE_HIDDEN_DIM: 64
|
||||
USE_DOT_PRODUCT_TOKEN_LOSS: True
|
||||
USE_LAYER_SCALE: True
|
||||
CLAMP_MIN_FOR_UNDERFLOW: True
|
||||
CLAMP_MAX_FOR_OVERFLOW: True
|
||||
CLAMP_BERTATTN_MIN_FOR_UNDERFLOW: True
|
||||
CLAMP_BERTATTN_MAX_FOR_OVERFLOW: True
|
||||
CLAMP_DOT_PRODUCT: True
|
||||
|
||||
TEST:
|
||||
EVAL_TASK: 'detection'
|
||||
DURING_TRAINING: False
|
||||
IMS_PER_BATCH: 8
|
||||
|
||||
DATASETS:
|
||||
|
||||
TRAIN: ("object365_grounding_train", )
|
||||
TEST: ("coco_2017_val", )
|
||||
|
||||
ONE_HOT: False
|
||||
FLICKR_COPY: 8 # 0.15 * 8 = ~1.2M
|
||||
MIXED_COPY: 4 # 0.6 * 4 = ~2.4M
|
||||
OBJECT365_COPY: 2 # 1.4 * 2 = ~2.8M
|
||||
VG_COPY: 3 # 0.4 * 3 = ~1.2M
|
||||
IN_COPY: 2 # 0.67 * 2 = ~1.33M
|
||||
OI_COPY: 1 # 2M * 1 = 2M
|
||||
|
||||
DISABLE_SHUFFLE: False
|
||||
ADD_DET_PROMPT: False
|
||||
RANDOM_SAMPLE_NEG: 85
|
||||
CONTROL_PROB: (0.0, 0.0, 0.5, 0.0)
|
||||
FURTHER_SCREEN: True
|
||||
CAPTION_CONF: 0.5
|
||||
CAPTION_NMS: -1.0
|
||||
CAPTION_MIN_BOX: 1
|
||||
|
||||
SEPARATION_TOKENS: ". "
|
||||
|
||||
PACK_RANDOM_CAPTION_NUMBER: 20
|
||||
NO_RANDOM_PACK_PROBABILITY: 0.4
|
||||
RANDOM_PACK_PROB: 0.5
|
||||
CAPTION_FORMAT_VERSION: "v2"
|
||||
|
||||
EXCLUDE_CROWD: True
|
||||
SPECIAL_SAFEGUARD_FOR_COCO_GROUNDING: True
|
||||
|
||||
INPUT:
|
||||
PIXEL_MEAN: [ 103.530, 116.280, 123.675 ]
|
||||
PIXEL_STD: [ 57.375, 57.120, 58.395 ]
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
|
||||
AUGMENT:
|
||||
MULT_MIN_SIZE_TRAIN: (480,560,640,720,800)
|
||||
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
|
||||
SOLVER:
|
||||
OPTIMIZER: ADAMW
|
||||
BASE_LR: 0.0001
|
||||
#### should be modified during fine-tuning #######
|
||||
GATE_LR: 0.0025
|
||||
QUERY_LR: 0.00001
|
||||
#################################################
|
||||
LANG_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.01
|
||||
WEIGHT_DECAY_SCHEDULE: True
|
||||
# STEPS: (0.67, 0.89)
|
||||
STEPS: (0.95,)
|
||||
# MAX_ITER: 1000000
|
||||
MAX_EPOCH: 1
|
||||
# IMS_PER_BATCH: 64
|
||||
IMS_PER_BATCH: 8
|
||||
WARMUP_ITERS: 2000
|
||||
WARMUP_FACTOR: 0.001
|
||||
|
||||
FIND_UNUSED_PARAMETERS: False
|
||||
|
||||
USE_AMP: True
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: 16.0
|
||||
TUNING_HIGHLEVEL_OVERRIDE: "vision_query"
|
||||
MAX_TO_KEEP: 4
|
||||
|
||||
CLIP_GRADIENTS:
|
||||
ENABLED: True
|
||||
CLIP_TYPE: "full_model"
|
||||
CLIP_VALUE: 1.0
|
||||
NORM_TYPE: 2.0
|
||||
|
||||
VISION_QUERY:
|
||||
ENABLED: True
|
||||
QUERY_BANK_PATH: 'MODEL/object365_query_5000_pool7_sel_large.pth'
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.4
|
||||
VISION_SCALE: 1.0
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
ADD_ADAPT_LAYER: False
|
||||
CONDITION_GATE: True
|
||||
NONLINEAR_GATE: True
|
||||
NO_CAT: True
|
||||
QUERY_ADDITION_NAME: '_L'
|
||||
|
|
@ -0,0 +1,140 @@
|
|||
# object365_vs_sel_mod2_scale1_drop04_adaptor_woconcate
|
||||
MODEL:
|
||||
META_ARCHITECTURE: "GeneralizedVLRCNN_New"
|
||||
WEIGHT: "MODEL/glip_tiny_model_o365_goldg_cc_sbu.pth"
|
||||
# WEIGHT: "MODEL/mq-glip-t" # debug
|
||||
RPN_ONLY: True
|
||||
RPN_ARCHITECTURE: "VLDYHEAD"
|
||||
|
||||
BACKBONE:
|
||||
CONV_BODY: "SWINT-FPN-RETINANET"
|
||||
OUT_CHANNELS: 256
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
|
||||
LANGUAGE_BACKBONE:
|
||||
FREEZE: False
|
||||
TOKENIZER_TYPE: "bert-base-uncased"
|
||||
MODEL_TYPE: "bert-base-uncased"
|
||||
# TOKENIZER_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased" # debug
|
||||
# MODEL_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased"
|
||||
MASK_SPECIAL: False
|
||||
|
||||
ROI_BOX_HEAD:
|
||||
POOLER_RESOLUTION: 7
|
||||
POOLER_SCALES: (0.125, 0.0625, 0.03125, 0.015625, 0.0078125)
|
||||
POOLER_SAMPLING_RATIO: 0
|
||||
|
||||
RPN:
|
||||
USE_FPN: True
|
||||
ANCHOR_SIZES: (64, 128, 256, 512, 1024)
|
||||
ANCHOR_STRIDE: (8, 16, 32, 64, 128)
|
||||
ASPECT_RATIOS: (1.0,)
|
||||
SCALES_PER_OCTAVE: 1
|
||||
|
||||
DYHEAD:
|
||||
CHANNELS: 256
|
||||
NUM_CONVS: 6
|
||||
USE_GN: True
|
||||
USE_DYRELU: True
|
||||
USE_DFCONV: True
|
||||
USE_DYFUSE: True
|
||||
TOPK: 9 # topk for selecting candidate positive samples from each level
|
||||
SCORE_AGG: "MEAN"
|
||||
LOG_SCALE: 0.0
|
||||
|
||||
FUSE_CONFIG:
|
||||
EARLY_FUSE_ON: True
|
||||
TYPE: "MHA-B"
|
||||
USE_CLASSIFICATION_LOSS: False
|
||||
USE_TOKEN_LOSS: False
|
||||
USE_CONTRASTIVE_ALIGN_LOSS: False
|
||||
CONTRASTIVE_HIDDEN_DIM: 64
|
||||
USE_DOT_PRODUCT_TOKEN_LOSS: True
|
||||
USE_FUSED_FEATURES_DOT_PRODUCT: True
|
||||
USE_LAYER_SCALE: True
|
||||
CLAMP_MIN_FOR_UNDERFLOW: True
|
||||
CLAMP_MAX_FOR_OVERFLOW: True
|
||||
CLAMP_BERTATTN_MIN_FOR_UNDERFLOW: True
|
||||
CLAMP_BERTATTN_MAX_FOR_OVERFLOW: True
|
||||
CLAMP_DOT_PRODUCT: True
|
||||
|
||||
USE_CHECKPOINT: False
|
||||
|
||||
TEST:
|
||||
EVAL_TASK: 'detection'
|
||||
DURING_TRAINING: False
|
||||
IMS_PER_BATCH: 8
|
||||
|
||||
# use for grounding model
|
||||
DATASETS:
|
||||
TRAIN: ("object365_grounding_train", )
|
||||
TEST: ("coco_2017_val", )
|
||||
DISABLE_SHUFFLE: False
|
||||
ADD_DET_PROMPT: False
|
||||
RANDOM_SAMPLE_NEG: 85
|
||||
# RANDOM_SAMPLE_NEG: 365
|
||||
CONTROL_PROB: (0.0, 0.0, 0.5, 0.0)
|
||||
|
||||
SEPARATION_TOKENS: ". "
|
||||
|
||||
EXCLUDE_CROWD: True
|
||||
SPECIAL_SAFEGUARD_FOR_COCO_GROUNDING: True
|
||||
|
||||
INPUT:
|
||||
PIXEL_MEAN: [ 103.530, 116.280, 123.675 ]
|
||||
PIXEL_STD: [ 57.375, 57.120, 58.395 ]
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
|
||||
AUGMENT:
|
||||
MULT_MIN_SIZE_TRAIN: (480,560,640,720,800)
|
||||
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
|
||||
SOLVER:
|
||||
OPTIMIZER: ADAMW
|
||||
BASE_LR: 0.0001
|
||||
#### should be modified during fine-tuning #######
|
||||
GATE_LR: 0.005
|
||||
QUERY_LR: 0.00001
|
||||
#################################################
|
||||
LANG_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.0001
|
||||
# STEPS: (0.67, 0.89)
|
||||
STEPS: (0.95,)
|
||||
# MAX_EPOCH: 10
|
||||
MAX_EPOCH: 1
|
||||
IMS_PER_BATCH: 16
|
||||
WARMUP_ITERS: 2000
|
||||
WARMUP_FACTOR: 0.001
|
||||
USE_AMP: True
|
||||
MODEL_EMA: 0.999
|
||||
FIND_UNUSED_PARAMETERS: False
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: 128.0
|
||||
TUNING_HIGHLEVEL_OVERRIDE: "vision_query"
|
||||
MAX_TO_KEEP: 4
|
||||
|
||||
|
||||
CLIP_GRADIENTS:
|
||||
ENABLED: True
|
||||
CLIP_TYPE: "full_model"
|
||||
CLIP_VALUE: 1.0
|
||||
NORM_TYPE: 2.0
|
||||
|
||||
VISION_QUERY:
|
||||
ENABLED: True
|
||||
QUERY_BANK_PATH: 'MODEL/object365_query_5000_sel_tiny.pth'
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.4
|
||||
VISION_SCALE: 1.0
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
ADD_ADAPT_LAYER: False
|
||||
CONDITION_GATE: True
|
||||
NONLINEAR_GATE: True
|
||||
NO_CAT: True
|
||||
|
||||
|
|
@ -0,0 +1,112 @@
|
|||
MODEL:
|
||||
WEIGHT: "MODEL/groundingdino_swint_ogc.pth"
|
||||
|
||||
BACKBONE:
|
||||
OUT_CHANNELS: 256
|
||||
|
||||
LANGUAGE_BACKBONE:
|
||||
FREEZE: False
|
||||
TOKENIZER_TYPE: "bert-base-uncased"
|
||||
MODEL_TYPE: "bert-base-uncased" # "roberta-base", "clip"
|
||||
# TOKENIZER_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased"
|
||||
# MODEL_TYPE: "MODEL/THIRD_PARTIES/bert-base-uncased" # "roberta-base", "clip"
|
||||
MASK_SPECIAL: False
|
||||
|
||||
ROI_BOX_HEAD:
|
||||
POOLER_RESOLUTION: 7
|
||||
POOLER_SCALES: (0.125, 0.0625, 0.03125, 0.015625)
|
||||
POOLER_SAMPLING_RATIO: 0
|
||||
|
||||
|
||||
TEST:
|
||||
EVAL_TASK: 'detection'
|
||||
DURING_TRAINING: False
|
||||
IMS_PER_BATCH: 8
|
||||
|
||||
# use for grounding model
|
||||
DATASETS:
|
||||
TRAIN: ("object365_grounding_train", )
|
||||
TEST: ("coco_2017_val", )
|
||||
DISABLE_SHUFFLE: False
|
||||
ADD_DET_PROMPT: False
|
||||
RANDOM_SAMPLE_NEG: 85
|
||||
# RANDOM_SAMPLE_NEG: 365
|
||||
CONTROL_PROB: (0.0, 0.0, 0.5, 0.0)
|
||||
|
||||
SEPARATION_TOKENS: ". "
|
||||
|
||||
EXCLUDE_CROWD: True
|
||||
SPECIAL_SAFEGUARD_FOR_COCO_GROUNDING: True
|
||||
SEP_AT_LAST: True
|
||||
ADD_NORMED_CXCY: True
|
||||
|
||||
INPUT:
|
||||
FORMAT: 'rgb'
|
||||
PIXEL_MEAN: [0.485, 0.456, 0.406]
|
||||
PIXEL_STD: [0.229, 0.224, 0.225]
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
|
||||
AUGMENT:
|
||||
MULT_MIN_SIZE_TRAIN: (480,560,640,720,800)
|
||||
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
|
||||
SOLVER:
|
||||
OPTIMIZER: ADAMW
|
||||
BASE_LR: 0.0001
|
||||
#### should be modified during fine-tuning #######
|
||||
GATE_LR: 0.005
|
||||
QUERY_LR: 0.00001
|
||||
#################################################
|
||||
LANG_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.0001
|
||||
# STEPS: (0.67, 0.89)
|
||||
STEPS: (0.95,)
|
||||
# MAX_EPOCH: 10
|
||||
MAX_EPOCH: 1
|
||||
IMS_PER_BATCH: 16
|
||||
WARMUP_ITERS: 2000
|
||||
WARMUP_FACTOR: 0.001
|
||||
USE_AMP: True
|
||||
MODEL_EMA: 0.999
|
||||
FIND_UNUSED_PARAMETERS: False
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: 32.0
|
||||
# TUNING_HIGHLEVEL_OVERRIDE: "vision_query"
|
||||
TUNING_HIGHLEVEL_OVERRIDE: "vision_query"
|
||||
MAX_TO_KEEP: 4
|
||||
|
||||
|
||||
CLIP_GRADIENTS:
|
||||
ENABLED: True
|
||||
CLIP_TYPE: "full_model"
|
||||
CLIP_VALUE: 1.0
|
||||
# CLIP_VALUE: 0.1
|
||||
NORM_TYPE: 2.0
|
||||
|
||||
VISION_QUERY:
|
||||
ENABLED: True
|
||||
QUERY_BANK_PATH: 'MODEL/object365_query_5000_pool7_sel_gd.pth'
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.4
|
||||
VISION_SCALE: 1.0
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
ADD_ADAPT_LAYER: False
|
||||
CONDITION_GATE: True
|
||||
NONLINEAR_GATE: True
|
||||
NO_CAT: True
|
||||
QUERY_ADDITION_NAME: '_groundingdino-T'
|
||||
|
||||
GROUNDINGDINO:
|
||||
enabled: True
|
||||
use_checkpoint: False
|
||||
use_transformer_ckpt: False
|
||||
text_encoder_type: 'bert-base-uncased'
|
||||
box_threshold: 0.05
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,44 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 81
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
# WEIGHT_DECAY: 0.05
|
||||
WEIGHT_DECAY: 0.0001
|
||||
WARMUP_ITERS: 2000
|
||||
# USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
TEST:
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
DATASETS:
|
||||
TRAIN: ("coco_grounding_train_for_obj365", )
|
||||
TEST: ("coco_2017_val", )
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: True
|
||||
FEW_SHOT: 5
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: MODEL/coco_query_5_sel.pth
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
MAX_QUERY_NUMBER: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_mini_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_mini_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel_large.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_mini_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel_gd.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,71 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
|
@ -0,0 +1,71 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel_large.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
|
@ -0,0 +1,73 @@
|
|||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: -1
|
||||
ATSS:
|
||||
NUM_CLASSES: 8 # these fields are not used; just a placeholder
|
||||
DETECTIONS_PER_IMG: 300
|
||||
FCOS:
|
||||
NUM_CLASSES: 8
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_BOX_HEAD:
|
||||
NUM_CLASSES: 8
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 8
|
||||
RETINANET:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
ROI_HEADS:
|
||||
DETECTIONS_PER_IMG: 300
|
||||
DYHEAD:
|
||||
NUM_CLASSES: 1204
|
||||
|
||||
DATASETS:
|
||||
REGISTER:
|
||||
lvis_evaluation_mini_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_v1_minival_inserted_image_name.json"
|
||||
lvis_evaluation_val:
|
||||
img_dir: "coco"
|
||||
ann_file: "coco/annotations/lvis_od_val.json"
|
||||
TRAIN: ("lvis_grounding_train_for_obj365", )
|
||||
TEST: ("lvis_evaluation_val",)
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
DISABLE_SHUFFLE: False
|
||||
FEW_SHOT: 5
|
||||
|
||||
SOLVER:
|
||||
STEPS: (0.67, 0.89)
|
||||
BASE_LR: 0.00001
|
||||
LANG_LR: 0.00001
|
||||
GATE_LR: 0.0001
|
||||
QUERY_LR: 0.00001
|
||||
WEIGHT_DECAY: 0.05
|
||||
WARMUP_ITERS: 20
|
||||
USE_AUTOSTEP: True
|
||||
|
||||
TEST_WITH_INFERENCE: True
|
||||
|
||||
CHECKPOINT_PERIOD: 99999999
|
||||
CHECKPOINT_PER_EPOCH: -1.0
|
||||
|
||||
INPUT:
|
||||
MIN_SIZE_TRAIN: 800
|
||||
MAX_SIZE_TRAIN: 1333
|
||||
MIN_SIZE_TEST: 800
|
||||
MAX_SIZE_TEST: 1333
|
||||
DATALOADER:
|
||||
SIZE_DIVISIBILITY: 32
|
||||
ASPECT_RATIO_GROUPING: False
|
||||
TEST:
|
||||
IMS_PER_BATCH: 1
|
||||
CHUNKED_EVALUATION: 40
|
||||
MDETR_STYLE_AGGREGATE_CLASS_NUM: 3000
|
||||
DURING_TRAINING: False
|
||||
EVAL_TASK: detection
|
||||
|
||||
VISION_QUERY:
|
||||
QUERY_BANK_PATH: 'MODEL/lvis_query_5_pool7_sel_groundingdino_tiny.pth'
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
RANDOM_KSHOT: False
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,41 @@
|
|||
TEST:
|
||||
EVAL_TASK: detection
|
||||
DURING_TRAINING: True
|
||||
DATASETS:
|
||||
TRAIN_DATASETNAME_SUFFIX: '_vision_query'
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
USE_CAPTION_PROMPT: True
|
||||
FEW_SHOT: 5
|
||||
USE_OVERRIDE_CATEGORY: True
|
||||
SHUFFLE_SEED: 3
|
||||
DISABLE_SHUFFLE: True
|
||||
SPECIAL_SAFEGUARD_FOR_COCO_GROUNDING: False
|
||||
DATALOADER:
|
||||
DISTRIBUTE_CHUNK_AMONG_NODE: False
|
||||
VISION_QUERY:
|
||||
MAX_QUERY_NUMBER: 5
|
||||
VISION_SCALE: 1.0
|
||||
PURE_TEXT_RATE: 0.
|
||||
TEXT_DROPOUT: 0.
|
||||
NUM_QUERY_PER_CLASS: 5
|
||||
QUERY_BANK_PATH: ""
|
||||
RANDOM_KSHOT: False
|
||||
NUM_TURNS: 3
|
||||
OFFLINE_WITH_ONLINE: True
|
||||
SOLVER:
|
||||
WEIGHT_DECAY: 0.05
|
||||
GATE_LR: 0.0001
|
||||
TEST_WITH_INFERENCE: True
|
||||
USE_AUTOSTEP: True
|
||||
SEED: 10
|
||||
STEP_PATIENCE: 3
|
||||
CHECKPOINT_PER_EPOCH: 1.0
|
||||
AUTO_TERMINATE_PATIENCE: 10
|
||||
MODEL_EMA: 0.0
|
||||
TUNING_HIGHLEVEL_OVERRIDE: full
|
||||
MAX_TO_KEEP: 2
|
||||
MODEL:
|
||||
BACKBONE:
|
||||
FREEZE_CONV_BODY_AT: 2
|
||||
GROUNDINGDINO:
|
||||
box_threshold: 0.08
|
||||
|
|
@ -0,0 +1,43 @@
|
|||
batch_size = 1
|
||||
modelname = "groundingdino"
|
||||
backbone = "swin_B_384_22k"
|
||||
position_embedding = "sine"
|
||||
pe_temperatureH = 20
|
||||
pe_temperatureW = 20
|
||||
return_interm_indices = [1, 2, 3]
|
||||
backbone_freeze_keywords = None
|
||||
enc_layers = 6
|
||||
dec_layers = 6
|
||||
pre_norm = False
|
||||
dim_feedforward = 2048
|
||||
hidden_dim = 256
|
||||
dropout = 0.0
|
||||
nheads = 8
|
||||
num_queries = 900
|
||||
query_dim = 4
|
||||
num_patterns = 0
|
||||
num_feature_levels = 4
|
||||
enc_n_points = 4
|
||||
dec_n_points = 4
|
||||
two_stage_type = "standard"
|
||||
two_stage_bbox_embed_share = False
|
||||
two_stage_class_embed_share = False
|
||||
transformer_activation = "relu"
|
||||
dec_pred_bbox_embed_share = True
|
||||
dn_box_noise_scale = 1.0
|
||||
dn_label_noise_ratio = 0.5
|
||||
dn_label_coef = 1.0
|
||||
dn_bbox_coef = 1.0
|
||||
embed_init_tgt = True
|
||||
dn_labelbook_size = 2000
|
||||
max_text_len = 256
|
||||
text_encoder_type = "bert-base-uncased"
|
||||
use_text_enhancer = True
|
||||
use_fusion_layer = True
|
||||
use_checkpoint = True
|
||||
use_transformer_ckpt = True
|
||||
use_text_cross_attention = True
|
||||
text_dropout = 0.0
|
||||
fusion_dropout = 0.0
|
||||
fusion_droppath = 0.1
|
||||
sub_sentence_present = True
|
||||
|
|
@ -0,0 +1,43 @@
|
|||
batch_size = 1
|
||||
modelname = "groundingdino"
|
||||
backbone = "swin_T_224_1k"
|
||||
position_embedding = "sine"
|
||||
pe_temperatureH = 20
|
||||
pe_temperatureW = 20
|
||||
return_interm_indices = [1, 2, 3]
|
||||
backbone_freeze_keywords = None
|
||||
enc_layers = 6
|
||||
dec_layers = 6
|
||||
pre_norm = False
|
||||
dim_feedforward = 2048
|
||||
hidden_dim = 256
|
||||
dropout = 0.0
|
||||
nheads = 8
|
||||
num_queries = 900
|
||||
query_dim = 4
|
||||
num_patterns = 0
|
||||
num_feature_levels = 4
|
||||
enc_n_points = 4
|
||||
dec_n_points = 4
|
||||
two_stage_type = "standard"
|
||||
two_stage_bbox_embed_share = False
|
||||
two_stage_class_embed_share = False
|
||||
transformer_activation = "relu"
|
||||
dec_pred_bbox_embed_share = True
|
||||
dn_box_noise_scale = 1.0
|
||||
dn_label_noise_ratio = 0.5
|
||||
dn_label_coef = 1.0
|
||||
dn_bbox_coef = 1.0
|
||||
embed_init_tgt = True
|
||||
dn_labelbook_size = 2000
|
||||
max_text_len = 256
|
||||
text_encoder_type = "bert-base-uncased"
|
||||
use_text_enhancer = True
|
||||
use_fusion_layer = True
|
||||
use_checkpoint = True
|
||||
use_transformer_ckpt = True
|
||||
use_text_cross_attention = True
|
||||
text_dropout = 0.0
|
||||
fusion_dropout = 0.0
|
||||
fusion_droppath = 0.1
|
||||
sub_sentence_present = True
|
||||
|
|
@ -0,0 +1,311 @@
|
|||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
"""
|
||||
Transforms and data augmentation for both image + bbox.
|
||||
"""
|
||||
import os
|
||||
import random
|
||||
|
||||
import PIL
|
||||
import torch
|
||||
import torchvision.transforms as T
|
||||
import torchvision.transforms.functional as F
|
||||
|
||||
from groundingdino_new.util.box_ops import box_xyxy_to_cxcywh
|
||||
from groundingdino_new.util.misc import interpolate
|
||||
|
||||
|
||||
def crop(image, target, region):
|
||||
cropped_image = F.crop(image, *region)
|
||||
|
||||
target = target.copy()
|
||||
i, j, h, w = region
|
||||
|
||||
# should we do something wrt the original size?
|
||||
target["size"] = torch.tensor([h, w])
|
||||
|
||||
fields = ["labels", "area", "iscrowd", "positive_map"]
|
||||
|
||||
if "boxes" in target:
|
||||
boxes = target["boxes"]
|
||||
max_size = torch.as_tensor([w, h], dtype=torch.float32)
|
||||
cropped_boxes = boxes - torch.as_tensor([j, i, j, i])
|
||||
cropped_boxes = torch.min(cropped_boxes.reshape(-1, 2, 2), max_size)
|
||||
cropped_boxes = cropped_boxes.clamp(min=0)
|
||||
area = (cropped_boxes[:, 1, :] - cropped_boxes[:, 0, :]).prod(dim=1)
|
||||
target["boxes"] = cropped_boxes.reshape(-1, 4)
|
||||
target["area"] = area
|
||||
fields.append("boxes")
|
||||
|
||||
if "masks" in target:
|
||||
# FIXME should we update the area here if there are no boxes?
|
||||
target["masks"] = target["masks"][:, i : i + h, j : j + w]
|
||||
fields.append("masks")
|
||||
|
||||
# remove elements for which the boxes or masks that have zero area
|
||||
if "boxes" in target or "masks" in target:
|
||||
# favor boxes selection when defining which elements to keep
|
||||
# this is compatible with previous implementation
|
||||
if "boxes" in target:
|
||||
cropped_boxes = target["boxes"].reshape(-1, 2, 2)
|
||||
keep = torch.all(cropped_boxes[:, 1, :] > cropped_boxes[:, 0, :], dim=1)
|
||||
else:
|
||||
keep = target["masks"].flatten(1).any(1)
|
||||
|
||||
for field in fields:
|
||||
if field in target:
|
||||
target[field] = target[field][keep]
|
||||
|
||||
if os.environ.get("IPDB_SHILONG_DEBUG", None) == "INFO":
|
||||
# for debug and visualization only.
|
||||
if "strings_positive" in target:
|
||||
target["strings_positive"] = [
|
||||
_i for _i, _j in zip(target["strings_positive"], keep) if _j
|
||||
]
|
||||
|
||||
return cropped_image, target
|
||||
|
||||
|
||||
def hflip(image, target):
|
||||
flipped_image = F.hflip(image)
|
||||
|
||||
w, h = image.size
|
||||
|
||||
target = target.copy()
|
||||
if "boxes" in target:
|
||||
boxes = target["boxes"]
|
||||
boxes = boxes[:, [2, 1, 0, 3]] * torch.as_tensor([-1, 1, -1, 1]) + torch.as_tensor(
|
||||
[w, 0, w, 0]
|
||||
)
|
||||
target["boxes"] = boxes
|
||||
|
||||
if "masks" in target:
|
||||
target["masks"] = target["masks"].flip(-1)
|
||||
|
||||
return flipped_image, target
|
||||
|
||||
|
||||
def resize(image, target, size, max_size=None):
|
||||
# size can be min_size (scalar) or (w, h) tuple
|
||||
|
||||
def get_size_with_aspect_ratio(image_size, size, max_size=None):
|
||||
w, h = image_size
|
||||
if max_size is not None:
|
||||
min_original_size = float(min((w, h)))
|
||||
max_original_size = float(max((w, h)))
|
||||
if max_original_size / min_original_size * size > max_size:
|
||||
size = int(round(max_size * min_original_size / max_original_size))
|
||||
|
||||
if (w <= h and w == size) or (h <= w and h == size):
|
||||
return (h, w)
|
||||
|
||||
if w < h:
|
||||
ow = size
|
||||
oh = int(size * h / w)
|
||||
else:
|
||||
oh = size
|
||||
ow = int(size * w / h)
|
||||
|
||||
return (oh, ow)
|
||||
|
||||
def get_size(image_size, size, max_size=None):
|
||||
if isinstance(size, (list, tuple)):
|
||||
return size[::-1]
|
||||
else:
|
||||
return get_size_with_aspect_ratio(image_size, size, max_size)
|
||||
|
||||
size = get_size(image.size, size, max_size)
|
||||
rescaled_image = F.resize(image, size)
|
||||
|
||||
if target is None:
|
||||
return rescaled_image, None
|
||||
|
||||
ratios = tuple(float(s) / float(s_orig) for s, s_orig in zip(rescaled_image.size, image.size))
|
||||
ratio_width, ratio_height = ratios
|
||||
|
||||
target = target.copy()
|
||||
if "boxes" in target:
|
||||
boxes = target["boxes"]
|
||||
scaled_boxes = boxes * torch.as_tensor(
|
||||
[ratio_width, ratio_height, ratio_width, ratio_height]
|
||||
)
|
||||
target["boxes"] = scaled_boxes
|
||||
|
||||
if "area" in target:
|
||||
area = target["area"]
|
||||
scaled_area = area * (ratio_width * ratio_height)
|
||||
target["area"] = scaled_area
|
||||
|
||||
h, w = size
|
||||
target["size"] = torch.tensor([h, w])
|
||||
|
||||
if "masks" in target:
|
||||
target["masks"] = (
|
||||
interpolate(target["masks"][:, None].float(), size, mode="nearest")[:, 0] > 0.5
|
||||
)
|
||||
|
||||
return rescaled_image, target
|
||||
|
||||
|
||||
def pad(image, target, padding):
|
||||
# assumes that we only pad on the bottom right corners
|
||||
padded_image = F.pad(image, (0, 0, padding[0], padding[1]))
|
||||
if target is None:
|
||||
return padded_image, None
|
||||
target = target.copy()
|
||||
# should we do something wrt the original size?
|
||||
target["size"] = torch.tensor(padded_image.size[::-1])
|
||||
if "masks" in target:
|
||||
target["masks"] = torch.nn.functional.pad(target["masks"], (0, padding[0], 0, padding[1]))
|
||||
return padded_image, target
|
||||
|
||||
|
||||
class ResizeDebug(object):
|
||||
def __init__(self, size):
|
||||
self.size = size
|
||||
|
||||
def __call__(self, img, target):
|
||||
return resize(img, target, self.size)
|
||||
|
||||
|
||||
class RandomCrop(object):
|
||||
def __init__(self, size):
|
||||
self.size = size
|
||||
|
||||
def __call__(self, img, target):
|
||||
region = T.RandomCrop.get_params(img, self.size)
|
||||
return crop(img, target, region)
|
||||
|
||||
|
||||
class RandomSizeCrop(object):
|
||||
def __init__(self, min_size: int, max_size: int, respect_boxes: bool = False):
|
||||
# respect_boxes: True to keep all boxes
|
||||
# False to tolerence box filter
|
||||
self.min_size = min_size
|
||||
self.max_size = max_size
|
||||
self.respect_boxes = respect_boxes
|
||||
|
||||
def __call__(self, img: PIL.Image.Image, target: dict):
|
||||
init_boxes = len(target["boxes"])
|
||||
max_patience = 10
|
||||
for i in range(max_patience):
|
||||
w = random.randint(self.min_size, min(img.width, self.max_size))
|
||||
h = random.randint(self.min_size, min(img.height, self.max_size))
|
||||
region = T.RandomCrop.get_params(img, [h, w])
|
||||
result_img, result_target = crop(img, target, region)
|
||||
if (
|
||||
not self.respect_boxes
|
||||
or len(result_target["boxes"]) == init_boxes
|
||||
or i == max_patience - 1
|
||||
):
|
||||
return result_img, result_target
|
||||
return result_img, result_target
|
||||
|
||||
|
||||
class CenterCrop(object):
|
||||
def __init__(self, size):
|
||||
self.size = size
|
||||
|
||||
def __call__(self, img, target):
|
||||
image_width, image_height = img.size
|
||||
crop_height, crop_width = self.size
|
||||
crop_top = int(round((image_height - crop_height) / 2.0))
|
||||
crop_left = int(round((image_width - crop_width) / 2.0))
|
||||
return crop(img, target, (crop_top, crop_left, crop_height, crop_width))
|
||||
|
||||
|
||||
class RandomHorizontalFlip(object):
|
||||
def __init__(self, p=0.5):
|
||||
self.p = p
|
||||
|
||||
def __call__(self, img, target):
|
||||
if random.random() < self.p:
|
||||
return hflip(img, target)
|
||||
return img, target
|
||||
|
||||
|
||||
class RandomResize(object):
|
||||
def __init__(self, sizes, max_size=None):
|
||||
assert isinstance(sizes, (list, tuple))
|
||||
self.sizes = sizes
|
||||
self.max_size = max_size
|
||||
|
||||
def __call__(self, img, target=None):
|
||||
size = random.choice(self.sizes)
|
||||
return resize(img, target, size, self.max_size)
|
||||
|
||||
|
||||
class RandomPad(object):
|
||||
def __init__(self, max_pad):
|
||||
self.max_pad = max_pad
|
||||
|
||||
def __call__(self, img, target):
|
||||
pad_x = random.randint(0, self.max_pad)
|
||||
pad_y = random.randint(0, self.max_pad)
|
||||
return pad(img, target, (pad_x, pad_y))
|
||||
|
||||
|
||||
class RandomSelect(object):
|
||||
"""
|
||||
Randomly selects between transforms1 and transforms2,
|
||||
with probability p for transforms1 and (1 - p) for transforms2
|
||||
"""
|
||||
|
||||
def __init__(self, transforms1, transforms2, p=0.5):
|
||||
self.transforms1 = transforms1
|
||||
self.transforms2 = transforms2
|
||||
self.p = p
|
||||
|
||||
def __call__(self, img, target):
|
||||
if random.random() < self.p:
|
||||
return self.transforms1(img, target)
|
||||
return self.transforms2(img, target)
|
||||
|
||||
|
||||
class ToTensor(object):
|
||||
def __call__(self, img, target):
|
||||
return F.to_tensor(img), target
|
||||
|
||||
|
||||
class RandomErasing(object):
|
||||
def __init__(self, *args, **kwargs):
|
||||
self.eraser = T.RandomErasing(*args, **kwargs)
|
||||
|
||||
def __call__(self, img, target):
|
||||
return self.eraser(img), target
|
||||
|
||||
|
||||
class Normalize(object):
|
||||
def __init__(self, mean, std):
|
||||
self.mean = mean
|
||||
self.std = std
|
||||
|
||||
def __call__(self, image, target=None):
|
||||
image = F.normalize(image, mean=self.mean, std=self.std)
|
||||
if target is None:
|
||||
return image, None
|
||||
target = target.copy()
|
||||
h, w = image.shape[-2:]
|
||||
if "boxes" in target:
|
||||
boxes = target["boxes"]
|
||||
boxes = box_xyxy_to_cxcywh(boxes)
|
||||
boxes = boxes / torch.tensor([w, h, w, h], dtype=torch.float32)
|
||||
target["boxes"] = boxes
|
||||
return image, target
|
||||
|
||||
|
||||
class Compose(object):
|
||||
def __init__(self, transforms):
|
||||
self.transforms = transforms
|
||||
|
||||
def __call__(self, image, target):
|
||||
for t in self.transforms:
|
||||
image, target = t(image, target)
|
||||
return image, target
|
||||
|
||||
def __repr__(self):
|
||||
format_string = self.__class__.__name__ + "("
|
||||
for t in self.transforms:
|
||||
format_string += "\n"
|
||||
format_string += " {0}".format(t)
|
||||
format_string += "\n)"
|
||||
return format_string
|
||||
|
|
@ -0,0 +1,15 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Conditional DETR
|
||||
# Copyright (c) 2021 Microsoft. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Copied from DETR (https://github.com/facebookresearch/detr)
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
from .groundingdino import build_groundingdino
|
||||
|
|
@ -0,0 +1 @@
|
|||
from .backbone import build_backbone
|
||||
|
|
@ -0,0 +1,221 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Conditional DETR
|
||||
# Copyright (c) 2021 Microsoft. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Copied from DETR (https://github.com/facebookresearch/detr)
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
"""
|
||||
Backbone modules.
|
||||
"""
|
||||
|
||||
from typing import Dict, List
|
||||
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
import torchvision
|
||||
from torch import nn
|
||||
from torchvision.models._utils import IntermediateLayerGetter
|
||||
|
||||
from groundingdino_new.util.misc import NestedTensor, clean_state_dict, is_main_process
|
||||
|
||||
from .position_encoding import build_position_encoding
|
||||
from .swin_transformer import build_swin_transformer
|
||||
|
||||
|
||||
class FrozenBatchNorm2d(torch.nn.Module):
|
||||
"""
|
||||
BatchNorm2d where the batch statistics and the affine parameters are fixed.
|
||||
|
||||
Copy-paste from torchvision.misc.ops with added eps before rqsrt,
|
||||
without which any other models than torchvision.models.resnet[18,34,50,101]
|
||||
produce nans.
|
||||
"""
|
||||
|
||||
def __init__(self, n):
|
||||
super(FrozenBatchNorm2d, self).__init__()
|
||||
self.register_buffer("weight", torch.ones(n))
|
||||
self.register_buffer("bias", torch.zeros(n))
|
||||
self.register_buffer("running_mean", torch.zeros(n))
|
||||
self.register_buffer("running_var", torch.ones(n))
|
||||
|
||||
def _load_from_state_dict(
|
||||
self, state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
|
||||
):
|
||||
num_batches_tracked_key = prefix + "num_batches_tracked"
|
||||
if num_batches_tracked_key in state_dict:
|
||||
del state_dict[num_batches_tracked_key]
|
||||
|
||||
super(FrozenBatchNorm2d, self)._load_from_state_dict(
|
||||
state_dict, prefix, local_metadata, strict, missing_keys, unexpected_keys, error_msgs
|
||||
)
|
||||
|
||||
def forward(self, x):
|
||||
# move reshapes to the beginning
|
||||
# to make it fuser-friendly
|
||||
w = self.weight.reshape(1, -1, 1, 1)
|
||||
b = self.bias.reshape(1, -1, 1, 1)
|
||||
rv = self.running_var.reshape(1, -1, 1, 1)
|
||||
rm = self.running_mean.reshape(1, -1, 1, 1)
|
||||
eps = 1e-5
|
||||
scale = w * (rv + eps).rsqrt()
|
||||
bias = b - rm * scale
|
||||
return x * scale + bias
|
||||
|
||||
|
||||
class BackboneBase(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
backbone: nn.Module,
|
||||
train_backbone: bool,
|
||||
num_channels: int,
|
||||
return_interm_indices: list,
|
||||
):
|
||||
super().__init__()
|
||||
for name, parameter in backbone.named_parameters():
|
||||
if (
|
||||
not train_backbone
|
||||
or "layer2" not in name
|
||||
and "layer3" not in name
|
||||
and "layer4" not in name
|
||||
):
|
||||
parameter.requires_grad_(False)
|
||||
|
||||
return_layers = {}
|
||||
for idx, layer_index in enumerate(return_interm_indices):
|
||||
return_layers.update(
|
||||
{"layer{}".format(5 - len(return_interm_indices) + idx): "{}".format(layer_index)}
|
||||
)
|
||||
|
||||
# if len:
|
||||
# if use_stage1_feature:
|
||||
# return_layers = {"layer1": "0", "layer2": "1", "layer3": "2", "layer4": "3"}
|
||||
# else:
|
||||
# return_layers = {"layer2": "0", "layer3": "1", "layer4": "2"}
|
||||
# else:
|
||||
# return_layers = {'layer4': "0"}
|
||||
self.body = IntermediateLayerGetter(backbone, return_layers=return_layers)
|
||||
self.num_channels = num_channels
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
xs = self.body(tensor_list.tensors)
|
||||
out: Dict[str, NestedTensor] = {}
|
||||
for name, x in xs.items():
|
||||
m = tensor_list.mask
|
||||
assert m is not None
|
||||
mask = F.interpolate(m[None].float(), size=x.shape[-2:]).to(torch.bool)[0]
|
||||
out[name] = NestedTensor(x, mask)
|
||||
# import ipdb; ipdb.set_trace()
|
||||
return out
|
||||
|
||||
|
||||
class Backbone(BackboneBase):
|
||||
"""ResNet backbone with frozen BatchNorm."""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
name: str,
|
||||
train_backbone: bool,
|
||||
dilation: bool,
|
||||
return_interm_indices: list,
|
||||
batch_norm=FrozenBatchNorm2d,
|
||||
):
|
||||
if name in ["resnet18", "resnet34", "resnet50", "resnet101"]:
|
||||
backbone = getattr(torchvision.models, name)(
|
||||
replace_stride_with_dilation=[False, False, dilation],
|
||||
pretrained=is_main_process(),
|
||||
norm_layer=batch_norm,
|
||||
)
|
||||
else:
|
||||
raise NotImplementedError("Why you can get here with name {}".format(name))
|
||||
# num_channels = 512 if name in ('resnet18', 'resnet34') else 2048
|
||||
assert name not in ("resnet18", "resnet34"), "Only resnet50 and resnet101 are available."
|
||||
assert return_interm_indices in [[0, 1, 2, 3], [1, 2, 3], [3]]
|
||||
num_channels_all = [256, 512, 1024, 2048]
|
||||
num_channels = num_channels_all[4 - len(return_interm_indices) :]
|
||||
super().__init__(backbone, train_backbone, num_channels, return_interm_indices)
|
||||
|
||||
|
||||
class Joiner(nn.Sequential):
|
||||
def __init__(self, backbone, position_embedding):
|
||||
super().__init__(backbone, position_embedding)
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
xs = self[0](tensor_list)
|
||||
out: List[NestedTensor] = []
|
||||
pos = []
|
||||
for name, x in xs.items():
|
||||
out.append(x)
|
||||
# position encoding
|
||||
pos.append(self[1](x).to(x.tensors.dtype))
|
||||
|
||||
return out, pos
|
||||
|
||||
|
||||
def build_backbone(args):
|
||||
"""
|
||||
Useful args:
|
||||
- backbone: backbone name
|
||||
- lr_backbone:
|
||||
- dilation
|
||||
- return_interm_indices: available: [0,1,2,3], [1,2,3], [3]
|
||||
- backbone_freeze_keywords:
|
||||
- use_checkpoint: for swin only for now
|
||||
|
||||
"""
|
||||
position_embedding = build_position_encoding(args)
|
||||
train_backbone = True
|
||||
if not train_backbone:
|
||||
raise ValueError("Please set lr_backbone > 0")
|
||||
return_interm_indices = args.return_interm_indices
|
||||
assert return_interm_indices in [[0, 1, 2, 3], [1, 2, 3], [3]]
|
||||
args.backbone_freeze_keywords
|
||||
use_checkpoint = getattr(args, "use_checkpoint", False)
|
||||
|
||||
if args.backbone in ["resnet50", "resnet101"]:
|
||||
backbone = Backbone(
|
||||
args.backbone,
|
||||
train_backbone,
|
||||
args.dilation,
|
||||
return_interm_indices,
|
||||
batch_norm=FrozenBatchNorm2d,
|
||||
)
|
||||
bb_num_channels = backbone.num_channels
|
||||
elif args.backbone in [
|
||||
"swin_T_224_1k",
|
||||
"swin_B_224_22k",
|
||||
"swin_B_384_22k",
|
||||
"swin_L_224_22k",
|
||||
"swin_L_384_22k",
|
||||
]:
|
||||
pretrain_img_size = int(args.backbone.split("_")[-2])
|
||||
backbone = build_swin_transformer(
|
||||
args.backbone,
|
||||
pretrain_img_size=pretrain_img_size,
|
||||
out_indices=tuple(return_interm_indices),
|
||||
dilation=False,
|
||||
use_checkpoint=use_checkpoint,
|
||||
)
|
||||
|
||||
bb_num_channels = backbone.num_features[4 - len(return_interm_indices) :]
|
||||
else:
|
||||
raise NotImplementedError("Unknown backbone {}".format(args.backbone))
|
||||
|
||||
assert len(bb_num_channels) == len(
|
||||
return_interm_indices
|
||||
), f"len(bb_num_channels) {len(bb_num_channels)} != len(return_interm_indices) {len(return_interm_indices)}"
|
||||
|
||||
model = Joiner(backbone, position_embedding)
|
||||
model.num_channels = bb_num_channels
|
||||
assert isinstance(
|
||||
bb_num_channels, List
|
||||
), "bb_num_channels is expected to be a List but {}".format(type(bb_num_channels))
|
||||
# import ipdb; ipdb.set_trace()
|
||||
return model
|
||||
|
|
@ -0,0 +1,186 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# DINO
|
||||
# Copyright (c) 2022 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Conditional DETR
|
||||
# Copyright (c) 2021 Microsoft. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Copied from DETR (https://github.com/facebookresearch/detr)
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
"""
|
||||
Various positional encodings for the transformer.
|
||||
"""
|
||||
import math
|
||||
|
||||
import torch
|
||||
from torch import nn
|
||||
|
||||
from groundingdino_new.util.misc import NestedTensor
|
||||
|
||||
|
||||
class PositionEmbeddingSine(nn.Module):
|
||||
"""
|
||||
This is a more standard version of the position embedding, very similar to the one
|
||||
used by the Attention is all you need paper, generalized to work on images.
|
||||
"""
|
||||
|
||||
def __init__(self, num_pos_feats=64, temperature=10000, normalize=False, scale=None):
|
||||
super().__init__()
|
||||
self.num_pos_feats = num_pos_feats
|
||||
self.temperature = temperature
|
||||
self.normalize = normalize
|
||||
if scale is not None and normalize is False:
|
||||
raise ValueError("normalize should be True if scale is passed")
|
||||
if scale is None:
|
||||
scale = 2 * math.pi
|
||||
self.scale = scale
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
x = tensor_list.tensors
|
||||
mask = tensor_list.mask
|
||||
assert mask is not None
|
||||
not_mask = ~mask
|
||||
y_embed = not_mask.cumsum(1, dtype=torch.float32)
|
||||
x_embed = not_mask.cumsum(2, dtype=torch.float32)
|
||||
if self.normalize:
|
||||
eps = 1e-6
|
||||
# if os.environ.get("SHILONG_AMP", None) == '1':
|
||||
# eps = 1e-4
|
||||
# else:
|
||||
# eps = 1e-6
|
||||
y_embed = y_embed / (y_embed[:, -1:, :] + eps) * self.scale
|
||||
x_embed = x_embed / (x_embed[:, :, -1:] + eps) * self.scale
|
||||
|
||||
dim_t = torch.arange(self.num_pos_feats, dtype=torch.float32, device=x.device)
|
||||
dim_t = self.temperature ** (2 * (dim_t // 2) / self.num_pos_feats)
|
||||
|
||||
pos_x = x_embed[:, :, :, None] / dim_t
|
||||
pos_y = y_embed[:, :, :, None] / dim_t
|
||||
pos_x = torch.stack(
|
||||
(pos_x[:, :, :, 0::2].sin(), pos_x[:, :, :, 1::2].cos()), dim=4
|
||||
).flatten(3)
|
||||
pos_y = torch.stack(
|
||||
(pos_y[:, :, :, 0::2].sin(), pos_y[:, :, :, 1::2].cos()), dim=4
|
||||
).flatten(3)
|
||||
pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2)
|
||||
return pos
|
||||
|
||||
|
||||
class PositionEmbeddingSineHW(nn.Module):
|
||||
"""
|
||||
This is a more standard version of the position embedding, very similar to the one
|
||||
used by the Attention is all you need paper, generalized to work on images.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self, num_pos_feats=64, temperatureH=10000, temperatureW=10000, normalize=False, scale=None
|
||||
):
|
||||
super().__init__()
|
||||
self.num_pos_feats = num_pos_feats
|
||||
self.temperatureH = temperatureH
|
||||
self.temperatureW = temperatureW
|
||||
self.normalize = normalize
|
||||
if scale is not None and normalize is False:
|
||||
raise ValueError("normalize should be True if scale is passed")
|
||||
if scale is None:
|
||||
scale = 2 * math.pi
|
||||
self.scale = scale
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
x = tensor_list.tensors
|
||||
mask = tensor_list.mask
|
||||
assert mask is not None
|
||||
not_mask = ~mask
|
||||
y_embed = not_mask.cumsum(1, dtype=torch.float32)
|
||||
x_embed = not_mask.cumsum(2, dtype=torch.float32)
|
||||
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
if self.normalize:
|
||||
eps = 1e-6
|
||||
y_embed = y_embed / (y_embed[:, -1:, :] + eps) * self.scale
|
||||
x_embed = x_embed / (x_embed[:, :, -1:] + eps) * self.scale
|
||||
|
||||
dim_tx = torch.arange(self.num_pos_feats, dtype=torch.float32, device=x.device)
|
||||
dim_tx = self.temperatureW ** (2 * (torch.div(dim_tx, 2, rounding_mode='floor')) / self.num_pos_feats)
|
||||
pos_x = x_embed[:, :, :, None] / dim_tx
|
||||
|
||||
dim_ty = torch.arange(self.num_pos_feats, dtype=torch.float32, device=x.device)
|
||||
dim_ty = self.temperatureH ** (2 * (torch.div(dim_ty, 2, rounding_mode='floor')) / self.num_pos_feats)
|
||||
pos_y = y_embed[:, :, :, None] / dim_ty
|
||||
|
||||
pos_x = torch.stack(
|
||||
(pos_x[:, :, :, 0::2].sin(), pos_x[:, :, :, 1::2].cos()), dim=4
|
||||
).flatten(3)
|
||||
pos_y = torch.stack(
|
||||
(pos_y[:, :, :, 0::2].sin(), pos_y[:, :, :, 1::2].cos()), dim=4
|
||||
).flatten(3)
|
||||
pos = torch.cat((pos_y, pos_x), dim=3).permute(0, 3, 1, 2)
|
||||
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
return pos
|
||||
|
||||
|
||||
class PositionEmbeddingLearned(nn.Module):
|
||||
"""
|
||||
Absolute pos embedding, learned.
|
||||
"""
|
||||
|
||||
def __init__(self, num_pos_feats=256):
|
||||
super().__init__()
|
||||
self.row_embed = nn.Embedding(50, num_pos_feats)
|
||||
self.col_embed = nn.Embedding(50, num_pos_feats)
|
||||
self.reset_parameters()
|
||||
|
||||
def reset_parameters(self):
|
||||
nn.init.uniform_(self.row_embed.weight)
|
||||
nn.init.uniform_(self.col_embed.weight)
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
x = tensor_list.tensors
|
||||
h, w = x.shape[-2:]
|
||||
i = torch.arange(w, device=x.device)
|
||||
j = torch.arange(h, device=x.device)
|
||||
x_emb = self.col_embed(i)
|
||||
y_emb = self.row_embed(j)
|
||||
pos = (
|
||||
torch.cat(
|
||||
[
|
||||
x_emb.unsqueeze(0).repeat(h, 1, 1),
|
||||
y_emb.unsqueeze(1).repeat(1, w, 1),
|
||||
],
|
||||
dim=-1,
|
||||
)
|
||||
.permute(2, 0, 1)
|
||||
.unsqueeze(0)
|
||||
.repeat(x.shape[0], 1, 1, 1)
|
||||
)
|
||||
return pos
|
||||
|
||||
|
||||
def build_position_encoding(args):
|
||||
N_steps = args.hidden_dim // 2
|
||||
if args.position_embedding in ("v2", "sine"):
|
||||
# TODO find a better way of exposing other arguments
|
||||
position_embedding = PositionEmbeddingSineHW(
|
||||
N_steps,
|
||||
temperatureH=args.pe_temperatureH,
|
||||
temperatureW=args.pe_temperatureW,
|
||||
normalize=True,
|
||||
)
|
||||
elif args.position_embedding in ("v3", "learned"):
|
||||
position_embedding = PositionEmbeddingLearned(N_steps)
|
||||
else:
|
||||
raise ValueError(f"not supported {args.position_embedding}")
|
||||
|
||||
return position_embedding
|
||||
|
|
@ -0,0 +1,802 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# DINO
|
||||
# Copyright (c) 2022 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# --------------------------------------------------------
|
||||
# modified from https://github.com/SwinTransformer/Swin-Transformer-Object-Detection/blob/master/mmdet/models/backbones/swin_transformer.py
|
||||
# --------------------------------------------------------
|
||||
|
||||
import numpy as np
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
import torch.utils.checkpoint as checkpoint
|
||||
from timm.models.layers import DropPath, to_2tuple, trunc_normal_
|
||||
|
||||
from groundingdino_new.util.misc import NestedTensor
|
||||
|
||||
|
||||
class Mlp(nn.Module):
|
||||
"""Multilayer perceptron."""
|
||||
|
||||
def __init__(
|
||||
self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.0
|
||||
):
|
||||
super().__init__()
|
||||
out_features = out_features or in_features
|
||||
hidden_features = hidden_features or in_features
|
||||
self.fc1 = nn.Linear(in_features, hidden_features)
|
||||
self.act = act_layer()
|
||||
self.fc2 = nn.Linear(hidden_features, out_features)
|
||||
self.drop = nn.Dropout(drop)
|
||||
|
||||
def forward(self, x):
|
||||
x = self.fc1(x)
|
||||
x = self.act(x)
|
||||
x = self.drop(x)
|
||||
x = self.fc2(x)
|
||||
x = self.drop(x)
|
||||
return x
|
||||
|
||||
|
||||
def window_partition(x, window_size):
|
||||
"""
|
||||
Args:
|
||||
x: (B, H, W, C)
|
||||
window_size (int): window size
|
||||
Returns:
|
||||
windows: (num_windows*B, window_size, window_size, C)
|
||||
"""
|
||||
B, H, W, C = x.shape
|
||||
x = x.view(B, H // window_size, window_size, W // window_size, window_size, C)
|
||||
windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C)
|
||||
return windows
|
||||
|
||||
|
||||
def window_reverse(windows, window_size, H, W):
|
||||
"""
|
||||
Args:
|
||||
windows: (num_windows*B, window_size, window_size, C)
|
||||
window_size (int): Window size
|
||||
H (int): Height of image
|
||||
W (int): Width of image
|
||||
Returns:
|
||||
x: (B, H, W, C)
|
||||
"""
|
||||
B = int(windows.shape[0] / (H * W / window_size / window_size))
|
||||
x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1)
|
||||
x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1)
|
||||
return x
|
||||
|
||||
|
||||
class WindowAttention(nn.Module):
|
||||
"""Window based multi-head self attention (W-MSA) module with relative position bias.
|
||||
It supports both of shifted and non-shifted window.
|
||||
Args:
|
||||
dim (int): Number of input channels.
|
||||
window_size (tuple[int]): The height and width of the window.
|
||||
num_heads (int): Number of attention heads.
|
||||
qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
|
||||
qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set
|
||||
attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0
|
||||
proj_drop (float, optional): Dropout ratio of output. Default: 0.0
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
dim,
|
||||
window_size,
|
||||
num_heads,
|
||||
qkv_bias=True,
|
||||
qk_scale=None,
|
||||
attn_drop=0.0,
|
||||
proj_drop=0.0,
|
||||
):
|
||||
|
||||
super().__init__()
|
||||
self.dim = dim
|
||||
self.window_size = window_size # Wh, Ww
|
||||
self.num_heads = num_heads
|
||||
head_dim = dim // num_heads
|
||||
self.scale = qk_scale or head_dim**-0.5
|
||||
|
||||
# define a parameter table of relative position bias
|
||||
self.relative_position_bias_table = nn.Parameter(
|
||||
torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)
|
||||
) # 2*Wh-1 * 2*Ww-1, nH
|
||||
|
||||
# get pair-wise relative position index for each token inside the window
|
||||
coords_h = torch.arange(self.window_size[0])
|
||||
coords_w = torch.arange(self.window_size[1])
|
||||
coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww
|
||||
coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww
|
||||
relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww
|
||||
relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2
|
||||
relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0
|
||||
relative_coords[:, :, 1] += self.window_size[1] - 1
|
||||
relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1
|
||||
relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww
|
||||
self.register_buffer("relative_position_index", relative_position_index)
|
||||
|
||||
self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias)
|
||||
self.attn_drop = nn.Dropout(attn_drop)
|
||||
self.proj = nn.Linear(dim, dim)
|
||||
self.proj_drop = nn.Dropout(proj_drop)
|
||||
|
||||
trunc_normal_(self.relative_position_bias_table, std=0.02)
|
||||
self.softmax = nn.Softmax(dim=-1)
|
||||
|
||||
def forward(self, x, mask=None):
|
||||
"""Forward function.
|
||||
Args:
|
||||
x: input features with shape of (num_windows*B, N, C)
|
||||
mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None
|
||||
"""
|
||||
B_, N, C = x.shape
|
||||
qkv = (
|
||||
self.qkv(x)
|
||||
.reshape(B_, N, 3, self.num_heads, C // self.num_heads)
|
||||
.permute(2, 0, 3, 1, 4)
|
||||
)
|
||||
q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple)
|
||||
|
||||
q = q * self.scale
|
||||
attn = q @ k.transpose(-2, -1)
|
||||
|
||||
relative_position_bias = self.relative_position_bias_table[
|
||||
self.relative_position_index.view(-1)
|
||||
].view(
|
||||
self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1
|
||||
) # Wh*Ww,Wh*Ww,nH
|
||||
relative_position_bias = relative_position_bias.permute(
|
||||
2, 0, 1
|
||||
).contiguous() # nH, Wh*Ww, Wh*Ww
|
||||
attn = attn + relative_position_bias.unsqueeze(0)
|
||||
|
||||
if mask is not None:
|
||||
nW = mask.shape[0]
|
||||
attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0)
|
||||
attn = attn.view(-1, self.num_heads, N, N)
|
||||
attn = self.softmax(attn)
|
||||
else:
|
||||
attn = self.softmax(attn)
|
||||
|
||||
attn = self.attn_drop(attn)
|
||||
|
||||
x = (attn @ v).transpose(1, 2).reshape(B_, N, C)
|
||||
x = self.proj(x)
|
||||
x = self.proj_drop(x)
|
||||
return x
|
||||
|
||||
|
||||
class SwinTransformerBlock(nn.Module):
|
||||
"""Swin Transformer Block.
|
||||
Args:
|
||||
dim (int): Number of input channels.
|
||||
num_heads (int): Number of attention heads.
|
||||
window_size (int): Window size.
|
||||
shift_size (int): Shift size for SW-MSA.
|
||||
mlp_ratio (float): Ratio of mlp hidden dim to embedding dim.
|
||||
qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
|
||||
qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
|
||||
drop (float, optional): Dropout rate. Default: 0.0
|
||||
attn_drop (float, optional): Attention dropout rate. Default: 0.0
|
||||
drop_path (float, optional): Stochastic depth rate. Default: 0.0
|
||||
act_layer (nn.Module, optional): Activation layer. Default: nn.GELU
|
||||
norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
dim,
|
||||
num_heads,
|
||||
window_size=7,
|
||||
shift_size=0,
|
||||
mlp_ratio=4.0,
|
||||
qkv_bias=True,
|
||||
qk_scale=None,
|
||||
drop=0.0,
|
||||
attn_drop=0.0,
|
||||
drop_path=0.0,
|
||||
act_layer=nn.GELU,
|
||||
norm_layer=nn.LayerNorm,
|
||||
):
|
||||
super().__init__()
|
||||
self.dim = dim
|
||||
self.num_heads = num_heads
|
||||
self.window_size = window_size
|
||||
self.shift_size = shift_size
|
||||
self.mlp_ratio = mlp_ratio
|
||||
assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size"
|
||||
|
||||
self.norm1 = norm_layer(dim)
|
||||
self.attn = WindowAttention(
|
||||
dim,
|
||||
window_size=to_2tuple(self.window_size),
|
||||
num_heads=num_heads,
|
||||
qkv_bias=qkv_bias,
|
||||
qk_scale=qk_scale,
|
||||
attn_drop=attn_drop,
|
||||
proj_drop=drop,
|
||||
)
|
||||
|
||||
self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
|
||||
self.norm2 = norm_layer(dim)
|
||||
mlp_hidden_dim = int(dim * mlp_ratio)
|
||||
self.mlp = Mlp(
|
||||
in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop
|
||||
)
|
||||
|
||||
self.H = None
|
||||
self.W = None
|
||||
|
||||
def forward(self, x, mask_matrix):
|
||||
"""Forward function.
|
||||
Args:
|
||||
x: Input feature, tensor size (B, H*W, C).
|
||||
H, W: Spatial resolution of the input feature.
|
||||
mask_matrix: Attention mask for cyclic shift.
|
||||
"""
|
||||
B, L, C = x.shape
|
||||
H, W = self.H, self.W
|
||||
assert L == H * W, "input feature has wrong size"
|
||||
|
||||
shortcut = x
|
||||
x = self.norm1(x)
|
||||
x = x.view(B, H, W, C)
|
||||
|
||||
# pad feature maps to multiples of window size
|
||||
pad_l = pad_t = 0
|
||||
pad_r = (self.window_size - W % self.window_size) % self.window_size
|
||||
pad_b = (self.window_size - H % self.window_size) % self.window_size
|
||||
x = F.pad(x, (0, 0, pad_l, pad_r, pad_t, pad_b))
|
||||
_, Hp, Wp, _ = x.shape
|
||||
|
||||
# cyclic shift
|
||||
if self.shift_size > 0:
|
||||
shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2))
|
||||
attn_mask = mask_matrix
|
||||
else:
|
||||
shifted_x = x
|
||||
attn_mask = None
|
||||
|
||||
# partition windows
|
||||
x_windows = window_partition(
|
||||
shifted_x, self.window_size
|
||||
) # nW*B, window_size, window_size, C
|
||||
x_windows = x_windows.view(
|
||||
-1, self.window_size * self.window_size, C
|
||||
) # nW*B, window_size*window_size, C
|
||||
|
||||
# W-MSA/SW-MSA
|
||||
attn_windows = self.attn(x_windows, mask=attn_mask) # nW*B, window_size*window_size, C
|
||||
|
||||
# merge windows
|
||||
attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C)
|
||||
shifted_x = window_reverse(attn_windows, self.window_size, Hp, Wp) # B H' W' C
|
||||
|
||||
# reverse cyclic shift
|
||||
if self.shift_size > 0:
|
||||
x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2))
|
||||
else:
|
||||
x = shifted_x
|
||||
|
||||
if pad_r > 0 or pad_b > 0:
|
||||
x = x[:, :H, :W, :].contiguous()
|
||||
|
||||
x = x.view(B, H * W, C)
|
||||
|
||||
# FFN
|
||||
x = shortcut + self.drop_path(x)
|
||||
x = x + self.drop_path(self.mlp(self.norm2(x)))
|
||||
|
||||
return x
|
||||
|
||||
|
||||
class PatchMerging(nn.Module):
|
||||
"""Patch Merging Layer
|
||||
Args:
|
||||
dim (int): Number of input channels.
|
||||
norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
|
||||
"""
|
||||
|
||||
def __init__(self, dim, norm_layer=nn.LayerNorm):
|
||||
super().__init__()
|
||||
self.dim = dim
|
||||
self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False)
|
||||
self.norm = norm_layer(4 * dim)
|
||||
|
||||
def forward(self, x, H, W):
|
||||
"""Forward function.
|
||||
Args:
|
||||
x: Input feature, tensor size (B, H*W, C).
|
||||
H, W: Spatial resolution of the input feature.
|
||||
"""
|
||||
B, L, C = x.shape
|
||||
assert L == H * W, "input feature has wrong size"
|
||||
|
||||
x = x.view(B, H, W, C)
|
||||
|
||||
# padding
|
||||
pad_input = (H % 2 == 1) or (W % 2 == 1)
|
||||
if pad_input:
|
||||
x = F.pad(x, (0, 0, 0, W % 2, 0, H % 2))
|
||||
|
||||
x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C
|
||||
x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C
|
||||
x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C
|
||||
x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C
|
||||
x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C
|
||||
x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C
|
||||
|
||||
x = self.norm(x)
|
||||
x = self.reduction(x)
|
||||
|
||||
return x
|
||||
|
||||
|
||||
class BasicLayer(nn.Module):
|
||||
"""A basic Swin Transformer layer for one stage.
|
||||
Args:
|
||||
dim (int): Number of feature channels
|
||||
depth (int): Depths of this stage.
|
||||
num_heads (int): Number of attention head.
|
||||
window_size (int): Local window size. Default: 7.
|
||||
mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
|
||||
qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True
|
||||
qk_scale (float | None, optional): Override default qk scale of head_dim ** -0.5 if set.
|
||||
drop (float, optional): Dropout rate. Default: 0.0
|
||||
attn_drop (float, optional): Attention dropout rate. Default: 0.0
|
||||
drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0
|
||||
norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm
|
||||
downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None
|
||||
use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
dim,
|
||||
depth,
|
||||
num_heads,
|
||||
window_size=7,
|
||||
mlp_ratio=4.0,
|
||||
qkv_bias=True,
|
||||
qk_scale=None,
|
||||
drop=0.0,
|
||||
attn_drop=0.0,
|
||||
drop_path=0.0,
|
||||
norm_layer=nn.LayerNorm,
|
||||
downsample=None,
|
||||
use_checkpoint=False,
|
||||
):
|
||||
super().__init__()
|
||||
self.window_size = window_size
|
||||
self.shift_size = window_size // 2
|
||||
self.depth = depth
|
||||
self.use_checkpoint = use_checkpoint
|
||||
|
||||
# build blocks
|
||||
self.blocks = nn.ModuleList(
|
||||
[
|
||||
SwinTransformerBlock(
|
||||
dim=dim,
|
||||
num_heads=num_heads,
|
||||
window_size=window_size,
|
||||
shift_size=0 if (i % 2 == 0) else window_size // 2,
|
||||
mlp_ratio=mlp_ratio,
|
||||
qkv_bias=qkv_bias,
|
||||
qk_scale=qk_scale,
|
||||
drop=drop,
|
||||
attn_drop=attn_drop,
|
||||
drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path,
|
||||
norm_layer=norm_layer,
|
||||
)
|
||||
for i in range(depth)
|
||||
]
|
||||
)
|
||||
|
||||
# patch merging layer
|
||||
if downsample is not None:
|
||||
self.downsample = downsample(dim=dim, norm_layer=norm_layer)
|
||||
else:
|
||||
self.downsample = None
|
||||
|
||||
def forward(self, x, H, W):
|
||||
"""Forward function.
|
||||
Args:
|
||||
x: Input feature, tensor size (B, H*W, C).
|
||||
H, W: Spatial resolution of the input feature.
|
||||
"""
|
||||
|
||||
# calculate attention mask for SW-MSA
|
||||
Hp = int(np.ceil(H / self.window_size)) * self.window_size
|
||||
Wp = int(np.ceil(W / self.window_size)) * self.window_size
|
||||
img_mask = torch.zeros((1, Hp, Wp, 1), device=x.device) # 1 Hp Wp 1
|
||||
h_slices = (
|
||||
slice(0, -self.window_size),
|
||||
slice(-self.window_size, -self.shift_size),
|
||||
slice(-self.shift_size, None),
|
||||
)
|
||||
w_slices = (
|
||||
slice(0, -self.window_size),
|
||||
slice(-self.window_size, -self.shift_size),
|
||||
slice(-self.shift_size, None),
|
||||
)
|
||||
cnt = 0
|
||||
for h in h_slices:
|
||||
for w in w_slices:
|
||||
img_mask[:, h, w, :] = cnt
|
||||
cnt += 1
|
||||
|
||||
mask_windows = window_partition(
|
||||
img_mask, self.window_size
|
||||
) # nW, window_size, window_size, 1
|
||||
mask_windows = mask_windows.view(-1, self.window_size * self.window_size)
|
||||
attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2)
|
||||
attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(
|
||||
attn_mask == 0, float(0.0)
|
||||
)
|
||||
|
||||
for blk in self.blocks:
|
||||
blk.H, blk.W = H, W
|
||||
if self.use_checkpoint:
|
||||
x = checkpoint.checkpoint(blk, x, attn_mask)
|
||||
else:
|
||||
x = blk(x, attn_mask)
|
||||
if self.downsample is not None:
|
||||
x_down = self.downsample(x, H, W)
|
||||
Wh, Ww = (H + 1) // 2, (W + 1) // 2
|
||||
return x, H, W, x_down, Wh, Ww
|
||||
else:
|
||||
return x, H, W, x, H, W
|
||||
|
||||
|
||||
class PatchEmbed(nn.Module):
|
||||
"""Image to Patch Embedding
|
||||
Args:
|
||||
patch_size (int): Patch token size. Default: 4.
|
||||
in_chans (int): Number of input image channels. Default: 3.
|
||||
embed_dim (int): Number of linear projection output channels. Default: 96.
|
||||
norm_layer (nn.Module, optional): Normalization layer. Default: None
|
||||
"""
|
||||
|
||||
def __init__(self, patch_size=4, in_chans=3, embed_dim=96, norm_layer=None):
|
||||
super().__init__()
|
||||
patch_size = to_2tuple(patch_size)
|
||||
self.patch_size = patch_size
|
||||
|
||||
self.in_chans = in_chans
|
||||
self.embed_dim = embed_dim
|
||||
|
||||
self.proj = nn.Conv2d(in_chans, embed_dim, kernel_size=patch_size, stride=patch_size)
|
||||
if norm_layer is not None:
|
||||
self.norm = norm_layer(embed_dim)
|
||||
else:
|
||||
self.norm = None
|
||||
|
||||
def forward(self, x):
|
||||
"""Forward function."""
|
||||
# padding
|
||||
_, _, H, W = x.size()
|
||||
if W % self.patch_size[1] != 0:
|
||||
x = F.pad(x, (0, self.patch_size[1] - W % self.patch_size[1]))
|
||||
if H % self.patch_size[0] != 0:
|
||||
x = F.pad(x, (0, 0, 0, self.patch_size[0] - H % self.patch_size[0]))
|
||||
|
||||
x = self.proj(x) # B C Wh Ww
|
||||
if self.norm is not None:
|
||||
Wh, Ww = x.size(2), x.size(3)
|
||||
x = x.flatten(2).transpose(1, 2)
|
||||
x = self.norm(x)
|
||||
x = x.transpose(1, 2).view(-1, self.embed_dim, Wh, Ww)
|
||||
|
||||
return x
|
||||
|
||||
|
||||
class SwinTransformer(nn.Module):
|
||||
"""Swin Transformer backbone.
|
||||
A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` -
|
||||
https://arxiv.org/pdf/2103.14030
|
||||
Args:
|
||||
pretrain_img_size (int): Input image size for training the pretrained model,
|
||||
used in absolute postion embedding. Default 224.
|
||||
patch_size (int | tuple(int)): Patch size. Default: 4.
|
||||
in_chans (int): Number of input image channels. Default: 3.
|
||||
embed_dim (int): Number of linear projection output channels. Default: 96.
|
||||
depths (tuple[int]): Depths of each Swin Transformer stage.
|
||||
num_heads (tuple[int]): Number of attention head of each stage.
|
||||
window_size (int): Window size. Default: 7.
|
||||
mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4.
|
||||
qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True
|
||||
qk_scale (float): Override default qk scale of head_dim ** -0.5 if set.
|
||||
drop_rate (float): Dropout rate.
|
||||
attn_drop_rate (float): Attention dropout rate. Default: 0.
|
||||
drop_path_rate (float): Stochastic depth rate. Default: 0.2.
|
||||
norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm.
|
||||
ape (bool): If True, add absolute position embedding to the patch embedding. Default: False.
|
||||
patch_norm (bool): If True, add normalization after patch embedding. Default: True.
|
||||
out_indices (Sequence[int]): Output from which stages.
|
||||
frozen_stages (int): Stages to be frozen (stop grad and set eval mode).
|
||||
-1 means not freezing any parameters.
|
||||
use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False.
|
||||
dilation (bool): if True, the output size if 16x downsample, ow 32x downsample.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
pretrain_img_size=224,
|
||||
patch_size=4,
|
||||
in_chans=3,
|
||||
embed_dim=96,
|
||||
depths=[2, 2, 6, 2],
|
||||
num_heads=[3, 6, 12, 24],
|
||||
window_size=7,
|
||||
mlp_ratio=4.0,
|
||||
qkv_bias=True,
|
||||
qk_scale=None,
|
||||
drop_rate=0.0,
|
||||
attn_drop_rate=0.0,
|
||||
drop_path_rate=0.2,
|
||||
norm_layer=nn.LayerNorm,
|
||||
ape=False,
|
||||
patch_norm=True,
|
||||
out_indices=(0, 1, 2, 3),
|
||||
frozen_stages=-1,
|
||||
dilation=False,
|
||||
use_checkpoint=False,
|
||||
):
|
||||
super().__init__()
|
||||
|
||||
self.pretrain_img_size = pretrain_img_size
|
||||
self.num_layers = len(depths)
|
||||
self.embed_dim = embed_dim
|
||||
self.ape = ape
|
||||
self.patch_norm = patch_norm
|
||||
self.out_indices = out_indices
|
||||
self.frozen_stages = frozen_stages
|
||||
self.dilation = dilation
|
||||
|
||||
# if use_checkpoint:
|
||||
# print("use_checkpoint!!!!!!!!!!!!!!!!!!!!!!!!")
|
||||
|
||||
# split image into non-overlapping patches
|
||||
self.patch_embed = PatchEmbed(
|
||||
patch_size=patch_size,
|
||||
in_chans=in_chans,
|
||||
embed_dim=embed_dim,
|
||||
norm_layer=norm_layer if self.patch_norm else None,
|
||||
)
|
||||
|
||||
# absolute position embedding
|
||||
if self.ape:
|
||||
pretrain_img_size = to_2tuple(pretrain_img_size)
|
||||
patch_size = to_2tuple(patch_size)
|
||||
patches_resolution = [
|
||||
pretrain_img_size[0] // patch_size[0],
|
||||
pretrain_img_size[1] // patch_size[1],
|
||||
]
|
||||
|
||||
self.absolute_pos_embed = nn.Parameter(
|
||||
torch.zeros(1, embed_dim, patches_resolution[0], patches_resolution[1])
|
||||
)
|
||||
trunc_normal_(self.absolute_pos_embed, std=0.02)
|
||||
|
||||
self.pos_drop = nn.Dropout(p=drop_rate)
|
||||
|
||||
# stochastic depth
|
||||
dpr = [
|
||||
x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))
|
||||
] # stochastic depth decay rule
|
||||
|
||||
# build layers
|
||||
self.layers = nn.ModuleList()
|
||||
# prepare downsample list
|
||||
downsamplelist = [PatchMerging for i in range(self.num_layers)]
|
||||
downsamplelist[-1] = None
|
||||
num_features = [int(embed_dim * 2**i) for i in range(self.num_layers)]
|
||||
if self.dilation:
|
||||
downsamplelist[-2] = None
|
||||
num_features[-1] = int(embed_dim * 2 ** (self.num_layers - 1)) // 2
|
||||
for i_layer in range(self.num_layers):
|
||||
layer = BasicLayer(
|
||||
# dim=int(embed_dim * 2 ** i_layer),
|
||||
dim=num_features[i_layer],
|
||||
depth=depths[i_layer],
|
||||
num_heads=num_heads[i_layer],
|
||||
window_size=window_size,
|
||||
mlp_ratio=mlp_ratio,
|
||||
qkv_bias=qkv_bias,
|
||||
qk_scale=qk_scale,
|
||||
drop=drop_rate,
|
||||
attn_drop=attn_drop_rate,
|
||||
drop_path=dpr[sum(depths[:i_layer]) : sum(depths[: i_layer + 1])],
|
||||
norm_layer=norm_layer,
|
||||
# downsample=PatchMerging if (i_layer < self.num_layers - 1) else None,
|
||||
downsample=downsamplelist[i_layer],
|
||||
use_checkpoint=use_checkpoint,
|
||||
)
|
||||
self.layers.append(layer)
|
||||
|
||||
# num_features = [int(embed_dim * 2 ** i) for i in range(self.num_layers)]
|
||||
self.num_features = num_features
|
||||
|
||||
# add a norm layer for each output
|
||||
for i_layer in out_indices:
|
||||
layer = norm_layer(num_features[i_layer])
|
||||
layer_name = f"norm{i_layer}"
|
||||
self.add_module(layer_name, layer)
|
||||
|
||||
self._freeze_stages()
|
||||
|
||||
def _freeze_stages(self):
|
||||
if self.frozen_stages >= 0:
|
||||
self.patch_embed.eval()
|
||||
for param in self.patch_embed.parameters():
|
||||
param.requires_grad = False
|
||||
|
||||
if self.frozen_stages >= 1 and self.ape:
|
||||
self.absolute_pos_embed.requires_grad = False
|
||||
|
||||
if self.frozen_stages >= 2:
|
||||
self.pos_drop.eval()
|
||||
for i in range(0, self.frozen_stages - 1):
|
||||
m = self.layers[i]
|
||||
m.eval()
|
||||
for param in m.parameters():
|
||||
param.requires_grad = False
|
||||
|
||||
# def init_weights(self, pretrained=None):
|
||||
# """Initialize the weights in backbone.
|
||||
# Args:
|
||||
# pretrained (str, optional): Path to pre-trained weights.
|
||||
# Defaults to None.
|
||||
# """
|
||||
|
||||
# def _init_weights(m):
|
||||
# if isinstance(m, nn.Linear):
|
||||
# trunc_normal_(m.weight, std=.02)
|
||||
# if isinstance(m, nn.Linear) and m.bias is not None:
|
||||
# nn.init.constant_(m.bias, 0)
|
||||
# elif isinstance(m, nn.LayerNorm):
|
||||
# nn.init.constant_(m.bias, 0)
|
||||
# nn.init.constant_(m.weight, 1.0)
|
||||
|
||||
# if isinstance(pretrained, str):
|
||||
# self.apply(_init_weights)
|
||||
# logger = get_root_logger()
|
||||
# load_checkpoint(self, pretrained, strict=False, logger=logger)
|
||||
# elif pretrained is None:
|
||||
# self.apply(_init_weights)
|
||||
# else:
|
||||
# raise TypeError('pretrained must be a str or None')
|
||||
|
||||
def forward_raw(self, x):
|
||||
"""Forward function."""
|
||||
x = self.patch_embed(x)
|
||||
|
||||
Wh, Ww = x.size(2), x.size(3)
|
||||
if self.ape:
|
||||
# interpolate the position embedding to the corresponding size
|
||||
absolute_pos_embed = F.interpolate(
|
||||
self.absolute_pos_embed, size=(Wh, Ww), mode="bicubic"
|
||||
)
|
||||
x = (x + absolute_pos_embed).flatten(2).transpose(1, 2) # B Wh*Ww C
|
||||
else:
|
||||
x = x.flatten(2).transpose(1, 2)
|
||||
x = self.pos_drop(x)
|
||||
|
||||
outs = []
|
||||
for i in range(self.num_layers):
|
||||
layer = self.layers[i]
|
||||
x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww)
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
if i in self.out_indices:
|
||||
norm_layer = getattr(self, f"norm{i}")
|
||||
x_out = norm_layer(x_out)
|
||||
|
||||
out = x_out.view(-1, H, W, self.num_features[i]).permute(0, 3, 1, 2).contiguous()
|
||||
outs.append(out)
|
||||
# in:
|
||||
# torch.Size([2, 3, 1024, 1024])
|
||||
# outs:
|
||||
# [torch.Size([2, 192, 256, 256]), torch.Size([2, 384, 128, 128]), \
|
||||
# torch.Size([2, 768, 64, 64]), torch.Size([2, 1536, 32, 32])]
|
||||
return tuple(outs)
|
||||
|
||||
def forward(self, tensor_list: NestedTensor):
|
||||
x = tensor_list.tensors
|
||||
|
||||
"""Forward function."""
|
||||
x = self.patch_embed(x)
|
||||
|
||||
Wh, Ww = x.size(2), x.size(3)
|
||||
if self.ape:
|
||||
# interpolate the position embedding to the corresponding size
|
||||
absolute_pos_embed = F.interpolate(
|
||||
self.absolute_pos_embed, size=(Wh, Ww), mode="bicubic"
|
||||
)
|
||||
x = (x + absolute_pos_embed).flatten(2).transpose(1, 2) # B Wh*Ww C
|
||||
else:
|
||||
x = x.flatten(2).transpose(1, 2)
|
||||
x = self.pos_drop(x)
|
||||
|
||||
outs = []
|
||||
for i in range(self.num_layers):
|
||||
layer = self.layers[i]
|
||||
x_out, H, W, x, Wh, Ww = layer(x, Wh, Ww)
|
||||
|
||||
if i in self.out_indices:
|
||||
norm_layer = getattr(self, f"norm{i}")
|
||||
x_out = norm_layer(x_out)
|
||||
|
||||
out = x_out.view(-1, H, W, self.num_features[i]).permute(0, 3, 1, 2).contiguous()
|
||||
outs.append(out)
|
||||
# in:
|
||||
# torch.Size([2, 3, 1024, 1024])
|
||||
# out:
|
||||
# [torch.Size([2, 192, 256, 256]), torch.Size([2, 384, 128, 128]), \
|
||||
# torch.Size([2, 768, 64, 64]), torch.Size([2, 1536, 32, 32])]
|
||||
|
||||
# collect for nesttensors
|
||||
outs_dict = {}
|
||||
for idx, out_i in enumerate(outs):
|
||||
m = tensor_list.mask
|
||||
assert m is not None
|
||||
mask = F.interpolate(m[None].float(), size=out_i.shape[-2:]).to(torch.bool)[0]
|
||||
outs_dict[idx] = NestedTensor(out_i, mask)
|
||||
|
||||
return outs_dict
|
||||
|
||||
def train(self, mode=True):
|
||||
"""Convert the model into training mode while keep layers freezed."""
|
||||
super(SwinTransformer, self).train(mode)
|
||||
self._freeze_stages()
|
||||
|
||||
|
||||
def build_swin_transformer(modelname, pretrain_img_size, **kw):
|
||||
assert modelname in [
|
||||
"swin_T_224_1k",
|
||||
"swin_B_224_22k",
|
||||
"swin_B_384_22k",
|
||||
"swin_L_224_22k",
|
||||
"swin_L_384_22k",
|
||||
]
|
||||
|
||||
model_para_dict = {
|
||||
"swin_T_224_1k": dict(
|
||||
embed_dim=96, depths=[2, 2, 6, 2], num_heads=[3, 6, 12, 24], window_size=7
|
||||
),
|
||||
"swin_B_224_22k": dict(
|
||||
embed_dim=128, depths=[2, 2, 18, 2], num_heads=[4, 8, 16, 32], window_size=7
|
||||
),
|
||||
"swin_B_384_22k": dict(
|
||||
embed_dim=128, depths=[2, 2, 18, 2], num_heads=[4, 8, 16, 32], window_size=12
|
||||
),
|
||||
"swin_L_224_22k": dict(
|
||||
embed_dim=192, depths=[2, 2, 18, 2], num_heads=[6, 12, 24, 48], window_size=7
|
||||
),
|
||||
"swin_L_384_22k": dict(
|
||||
embed_dim=192, depths=[2, 2, 18, 2], num_heads=[6, 12, 24, 48], window_size=12
|
||||
),
|
||||
}
|
||||
kw_cgf = model_para_dict[modelname]
|
||||
kw_cgf.update(kw)
|
||||
model = SwinTransformer(pretrain_img_size=pretrain_img_size, **kw_cgf)
|
||||
return model
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
model = build_swin_transformer("swin_L_384_22k", 384, dilation=True)
|
||||
x = torch.rand(2, 3, 1024, 1024)
|
||||
y = model.forward_raw(x)
|
||||
import ipdb
|
||||
|
||||
ipdb.set_trace()
|
||||
x = torch.rand(2, 3, 384, 384)
|
||||
y = model.forward_raw(x)
|
||||
|
|
@ -0,0 +1,320 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
import torch.utils.checkpoint as checkpoint
|
||||
from torch import Tensor, nn
|
||||
from torchvision.ops.boxes import nms
|
||||
from transformers import BertConfig, BertModel, BertPreTrainedModel
|
||||
from transformers.modeling_outputs import BaseModelOutputWithPoolingAndCrossAttentions
|
||||
|
||||
|
||||
def exists(val):
|
||||
if val is not None:
|
||||
if len(val) > 0:
|
||||
return True
|
||||
else:
|
||||
return False
|
||||
else:
|
||||
return False
|
||||
|
||||
class BertModelWarper(nn.Module):
|
||||
def __init__(self, bert_model):
|
||||
super().__init__()
|
||||
# self.bert = bert_modelc
|
||||
|
||||
self.config = bert_model.config
|
||||
self.embeddings = bert_model.embeddings
|
||||
self.encoder = bert_model.encoder
|
||||
self.pooler = bert_model.pooler
|
||||
try:
|
||||
self.pre_select = bert_model.pre_select
|
||||
except:
|
||||
pass
|
||||
try:
|
||||
self.cfg = bert_model.cfg
|
||||
except:
|
||||
pass
|
||||
|
||||
self.get_extended_attention_mask = bert_model.get_extended_attention_mask
|
||||
self.invert_attention_mask = bert_model.invert_attention_mask
|
||||
self.get_head_mask = bert_model.get_head_mask
|
||||
|
||||
def get_gate_value(self):
|
||||
attn_gates=[]
|
||||
ff_gates=[]
|
||||
for blk in self.encoder.qv_layer:
|
||||
# try:
|
||||
if not self.cfg.VISION_QUERY.CONDITION_GATE:
|
||||
attn_gates.append(blk.attn_gate)
|
||||
# except:
|
||||
# pass
|
||||
ff_gates.append(blk.ff_gate)
|
||||
return {'attn_gates': attn_gates, 'ffn_gates': ff_gates}
|
||||
|
||||
def forward(
|
||||
self,
|
||||
input_ids=None,
|
||||
attention_mask=None,
|
||||
token_type_ids=None,
|
||||
position_ids=None,
|
||||
head_mask=None,
|
||||
inputs_embeds=None,
|
||||
encoder_hidden_states=None,
|
||||
encoder_attention_mask=None,
|
||||
past_key_values=None,
|
||||
use_cache=None,
|
||||
output_attentions=None,
|
||||
output_hidden_states=None,
|
||||
return_dict=None,
|
||||
vision = None, # (batch, vision, dim)
|
||||
images = None, # (batch, image, dim)
|
||||
vision_attention_mask = None,
|
||||
batched_pos_category_map = None,
|
||||
):
|
||||
r"""
|
||||
encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`):
|
||||
Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if
|
||||
the model is configured as a decoder.
|
||||
encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`):
|
||||
Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in
|
||||
the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``:
|
||||
|
||||
- 1 for tokens that are **not masked**,
|
||||
- 0 for tokens that are **masked**.
|
||||
past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`):
|
||||
Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding.
|
||||
|
||||
If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids`
|
||||
(those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)`
|
||||
instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`.
|
||||
use_cache (:obj:`bool`, `optional`):
|
||||
If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up
|
||||
decoding (see :obj:`past_key_values`).
|
||||
"""
|
||||
output_attentions = (
|
||||
output_attentions if output_attentions is not None else self.config.output_attentions
|
||||
)
|
||||
output_hidden_states = (
|
||||
output_hidden_states
|
||||
if output_hidden_states is not None
|
||||
else self.config.output_hidden_states
|
||||
)
|
||||
return_dict = return_dict if return_dict is not None else self.config.use_return_dict
|
||||
|
||||
if self.config.is_decoder:
|
||||
use_cache = use_cache if use_cache is not None else self.config.use_cache
|
||||
else:
|
||||
use_cache = False
|
||||
|
||||
if input_ids is not None and inputs_embeds is not None:
|
||||
raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
|
||||
elif input_ids is not None:
|
||||
input_shape = input_ids.size()
|
||||
batch_size, seq_length = input_shape
|
||||
elif inputs_embeds is not None:
|
||||
input_shape = inputs_embeds.size()[:-1]
|
||||
batch_size, seq_length = input_shape
|
||||
else:
|
||||
raise ValueError("You have to specify either input_ids or inputs_embeds")
|
||||
|
||||
device = input_ids.device if input_ids is not None else inputs_embeds.device
|
||||
|
||||
# past_key_values_length
|
||||
past_key_values_length = (
|
||||
past_key_values[0][0].shape[2] if past_key_values is not None else 0
|
||||
)
|
||||
|
||||
if attention_mask is None:
|
||||
attention_mask = torch.ones(
|
||||
((batch_size, seq_length + past_key_values_length)), device=device
|
||||
)
|
||||
if token_type_ids is None:
|
||||
token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device)
|
||||
|
||||
# We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length]
|
||||
# ourselves in which case we just need to make it broadcastable to all heads.
|
||||
extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(
|
||||
attention_mask, input_shape, device
|
||||
)
|
||||
|
||||
# If a 2D or 3D attention mask is provided for the cross-attention
|
||||
# we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length]
|
||||
if self.config.is_decoder and encoder_hidden_states is not None:
|
||||
encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size()
|
||||
encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length)
|
||||
if encoder_attention_mask is None:
|
||||
encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device)
|
||||
encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask)
|
||||
else:
|
||||
encoder_extended_attention_mask = None
|
||||
# if os.environ.get('IPDB_SHILONG_DEBUG', None) == 'INFO':
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
# Prepare head mask if needed
|
||||
# 1.0 in head_mask indicate we keep the head
|
||||
# attention_probs has shape bsz x n_heads x N x N
|
||||
# input head_mask has shape [num_heads] or [num_hidden_layers x num_heads]
|
||||
# and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length]
|
||||
head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
|
||||
|
||||
embedding_output = self.embeddings(
|
||||
input_ids=input_ids,
|
||||
position_ids=position_ids,
|
||||
token_type_ids=token_type_ids,
|
||||
inputs_embeds=inputs_embeds,
|
||||
past_key_values_length=past_key_values_length,
|
||||
)
|
||||
|
||||
augmented_vision = None
|
||||
if (exists(images) and exists(vision)):
|
||||
vision = self.pre_select(vision, images)['vision']
|
||||
augmented_vision = vision
|
||||
|
||||
encoder_outputs = self.encoder(
|
||||
embedding_output,
|
||||
attention_mask=extended_attention_mask,
|
||||
head_mask=head_mask,
|
||||
encoder_hidden_states=encoder_hidden_states,
|
||||
encoder_attention_mask=encoder_extended_attention_mask,
|
||||
past_key_values=past_key_values,
|
||||
use_cache=use_cache,
|
||||
output_attentions=output_attentions,
|
||||
output_hidden_states=output_hidden_states,
|
||||
return_dict=return_dict,
|
||||
vision=vision,
|
||||
vision_attention_mask=vision_attention_mask,
|
||||
batched_pos_category_map=batched_pos_category_map,
|
||||
)
|
||||
sequence_output = encoder_outputs[0]
|
||||
pooled_output = self.pooler(sequence_output) if self.pooler is not None else None
|
||||
|
||||
if not return_dict:
|
||||
return (sequence_output, pooled_output) + encoder_outputs[1:]
|
||||
|
||||
out = BaseModelOutputWithPoolingAndCrossAttentions(
|
||||
last_hidden_state=sequence_output,
|
||||
pooler_output=pooled_output,
|
||||
past_key_values=encoder_outputs.past_key_values,
|
||||
hidden_states=encoder_outputs.hidden_states,
|
||||
attentions=encoder_outputs.attentions,
|
||||
cross_attentions=encoder_outputs.cross_attentions,
|
||||
)
|
||||
|
||||
out['vision_query_gates'] = self.get_gate_value()
|
||||
if self.cfg.VISION_QUERY.QUERY_FUSION:
|
||||
out['augmented_vision'] = augmented_vision
|
||||
out['vision_attention_mask'] = vision_attention_mask
|
||||
return out
|
||||
|
||||
|
||||
class TextEncoderShell(nn.Module):
|
||||
def __init__(self, text_encoder):
|
||||
super().__init__()
|
||||
self.text_encoder = text_encoder
|
||||
self.config = self.text_encoder.config
|
||||
|
||||
def forward(self, **kw):
|
||||
# feed into text encoder
|
||||
return self.text_encoder(**kw)
|
||||
|
||||
|
||||
def generate_masks_with_special_tokens(tokenized, special_tokens_list, tokenizer):
|
||||
"""Generate attention mask between each pair of special tokens
|
||||
Args:
|
||||
input_ids (torch.Tensor): input ids. Shape: [bs, num_token]
|
||||
special_tokens_mask (list): special tokens mask.
|
||||
Returns:
|
||||
torch.Tensor: attention mask between each special tokens.
|
||||
"""
|
||||
input_ids = tokenized["input_ids"]
|
||||
bs, num_token = input_ids.shape
|
||||
# special_tokens_mask: bs, num_token. 1 for special tokens. 0 for normal tokens
|
||||
special_tokens_mask = torch.zeros((bs, num_token), device=input_ids.device).bool()
|
||||
for special_token in special_tokens_list:
|
||||
special_tokens_mask |= input_ids == special_token
|
||||
|
||||
# idxs: each row is a list of indices of special tokens
|
||||
idxs = torch.nonzero(special_tokens_mask)
|
||||
|
||||
# generate attention mask and positional ids
|
||||
attention_mask = (
|
||||
torch.eye(num_token, device=input_ids.device).bool().unsqueeze(0).repeat(bs, 1, 1)
|
||||
)
|
||||
position_ids = torch.zeros((bs, num_token), device=input_ids.device)
|
||||
previous_col = 0
|
||||
for i in range(idxs.shape[0]):
|
||||
row, col = idxs[i]
|
||||
if (col == 0) or (col == num_token - 1):
|
||||
attention_mask[row, col, col] = True
|
||||
position_ids[row, col] = 0
|
||||
else:
|
||||
attention_mask[row, previous_col + 1 : col + 1, previous_col + 1 : col + 1] = True
|
||||
position_ids[row, previous_col + 1 : col + 1] = torch.arange(
|
||||
0, col - previous_col, device=input_ids.device
|
||||
)
|
||||
|
||||
previous_col = col
|
||||
|
||||
# # padding mask
|
||||
# padding_mask = tokenized['attention_mask']
|
||||
# attention_mask = attention_mask & padding_mask.unsqueeze(1).bool() & padding_mask.unsqueeze(2).bool()
|
||||
|
||||
return attention_mask, position_ids.to(torch.long)
|
||||
|
||||
|
||||
def generate_masks_with_special_tokens_and_transfer_map(tokenized, special_tokens_list, tokenizer):
|
||||
"""Generate attention mask between each pair of special tokens
|
||||
Args:
|
||||
input_ids (torch.Tensor): input ids. Shape: [bs, num_token]
|
||||
special_tokens_mask (list): special tokens mask.
|
||||
Returns:
|
||||
torch.Tensor: attention mask between each special tokens.
|
||||
"""
|
||||
input_ids = tokenized["input_ids"]
|
||||
bs, num_token = input_ids.shape
|
||||
# special_tokens_mask: bs, num_token. 1 for special tokens. 0 for normal tokens
|
||||
special_tokens_mask = torch.zeros((bs, num_token), device=input_ids.device).bool()
|
||||
for special_token in special_tokens_list:
|
||||
special_tokens_mask |= input_ids == special_token
|
||||
|
||||
# idxs: each row is a list of indices of special tokens
|
||||
idxs = torch.nonzero(special_tokens_mask)
|
||||
|
||||
# generate attention mask and positional ids
|
||||
attention_mask = (
|
||||
torch.eye(num_token, device=input_ids.device).bool().unsqueeze(0).repeat(bs, 1, 1)
|
||||
)
|
||||
position_ids = torch.zeros((bs, num_token), device=input_ids.device)
|
||||
cate_to_token_mask_list = [[] for _ in range(bs)]
|
||||
previous_col = 0
|
||||
for i in range(idxs.shape[0]):
|
||||
row, col = idxs[i]
|
||||
if (col == 0) or (col == num_token - 1):
|
||||
attention_mask[row, col, col] = True
|
||||
position_ids[row, col] = 0
|
||||
else:
|
||||
attention_mask[row, previous_col + 1 : col + 1, previous_col + 1 : col + 1] = True
|
||||
position_ids[row, previous_col + 1 : col + 1] = torch.arange(
|
||||
0, col - previous_col, device=input_ids.device
|
||||
)
|
||||
c2t_maski = torch.zeros((num_token), device=input_ids.device).bool()
|
||||
c2t_maski[previous_col + 1 : col] = True
|
||||
cate_to_token_mask_list[row].append(c2t_maski)
|
||||
previous_col = col
|
||||
|
||||
cate_to_token_mask_list = [
|
||||
torch.stack(cate_to_token_mask_listi, dim=0)
|
||||
for cate_to_token_mask_listi in cate_to_token_mask_list
|
||||
]
|
||||
|
||||
# # padding mask
|
||||
# padding_mask = tokenized['attention_mask']
|
||||
# attention_mask = attention_mask & padding_mask.unsqueeze(1).bool() & padding_mask.unsqueeze(2).bool()
|
||||
|
||||
return attention_mask, position_ids.to(torch.long), cate_to_token_mask_list
|
||||
|
|
@ -0,0 +1,64 @@
|
|||
/*!
|
||||
**************************************************************************************************
|
||||
* Deformable DETR
|
||||
* Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
**************************************************************************************************
|
||||
* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
|
||||
**************************************************************************************************
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
|
||||
#include "ms_deform_attn_cpu.h"
|
||||
|
||||
#ifdef WITH_CUDA
|
||||
#include "ms_deform_attn_cuda.h"
|
||||
#endif
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
at::Tensor
|
||||
ms_deform_attn_forward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const int im2col_step)
|
||||
{
|
||||
if (value.type().is_cuda())
|
||||
{
|
||||
#ifdef WITH_CUDA
|
||||
return ms_deform_attn_cuda_forward(
|
||||
value, spatial_shapes, level_start_index, sampling_loc, attn_weight, im2col_step);
|
||||
#else
|
||||
AT_ERROR("Not compiled with GPU support");
|
||||
#endif
|
||||
}
|
||||
AT_ERROR("Not implemented on the CPU");
|
||||
}
|
||||
|
||||
std::vector<at::Tensor>
|
||||
ms_deform_attn_backward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const at::Tensor &grad_output,
|
||||
const int im2col_step)
|
||||
{
|
||||
if (value.type().is_cuda())
|
||||
{
|
||||
#ifdef WITH_CUDA
|
||||
return ms_deform_attn_cuda_backward(
|
||||
value, spatial_shapes, level_start_index, sampling_loc, attn_weight, grad_output, im2col_step);
|
||||
#else
|
||||
AT_ERROR("Not compiled with GPU support");
|
||||
#endif
|
||||
}
|
||||
AT_ERROR("Not implemented on the CPU");
|
||||
}
|
||||
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,43 @@
|
|||
/*!
|
||||
**************************************************************************************************
|
||||
* Deformable DETR
|
||||
* Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
**************************************************************************************************
|
||||
* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
|
||||
**************************************************************************************************
|
||||
*/
|
||||
|
||||
#include <vector>
|
||||
|
||||
#include <ATen/ATen.h>
|
||||
#include <ATen/cuda/CUDAContext.h>
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
at::Tensor
|
||||
ms_deform_attn_cpu_forward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const int im2col_step)
|
||||
{
|
||||
AT_ERROR("Not implement on cpu");
|
||||
}
|
||||
|
||||
std::vector<at::Tensor>
|
||||
ms_deform_attn_cpu_backward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const at::Tensor &grad_output,
|
||||
const int im2col_step)
|
||||
{
|
||||
AT_ERROR("Not implement on cpu");
|
||||
}
|
||||
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,35 @@
|
|||
/*!
|
||||
**************************************************************************************************
|
||||
* Deformable DETR
|
||||
* Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
**************************************************************************************************
|
||||
* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
|
||||
**************************************************************************************************
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
#include <torch/extension.h>
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
at::Tensor
|
||||
ms_deform_attn_cpu_forward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const int im2col_step);
|
||||
|
||||
std::vector<at::Tensor>
|
||||
ms_deform_attn_cpu_backward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const at::Tensor &grad_output,
|
||||
const int im2col_step);
|
||||
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,156 @@
|
|||
/*!
|
||||
**************************************************************************************************
|
||||
* Deformable DETR
|
||||
* Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
**************************************************************************************************
|
||||
* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
|
||||
**************************************************************************************************
|
||||
*/
|
||||
|
||||
#include <vector>
|
||||
#include "ms_deform_im2col_cuda.cuh"
|
||||
|
||||
#include <ATen/ATen.h>
|
||||
#include <ATen/cuda/CUDAContext.h>
|
||||
#include <cuda.h>
|
||||
#include <cuda_runtime.h>
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
at::Tensor ms_deform_attn_cuda_forward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const int im2col_step)
|
||||
{
|
||||
AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
|
||||
AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
|
||||
AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
|
||||
AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
|
||||
AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
|
||||
|
||||
AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
|
||||
AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
|
||||
AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
|
||||
AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
|
||||
AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
|
||||
|
||||
const int batch = value.size(0);
|
||||
const int spatial_size = value.size(1);
|
||||
const int num_heads = value.size(2);
|
||||
const int channels = value.size(3);
|
||||
|
||||
const int num_levels = spatial_shapes.size(0);
|
||||
|
||||
const int num_query = sampling_loc.size(1);
|
||||
const int num_point = sampling_loc.size(4);
|
||||
|
||||
const int im2col_step_ = std::min(batch, im2col_step);
|
||||
|
||||
AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
|
||||
|
||||
auto output = at::zeros({batch, num_query, num_heads, channels}, value.options());
|
||||
|
||||
const int batch_n = im2col_step_;
|
||||
auto output_n = output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
|
||||
auto per_value_size = spatial_size * num_heads * channels;
|
||||
auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
|
||||
auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
|
||||
for (int n = 0; n < batch/im2col_step_; ++n)
|
||||
{
|
||||
auto columns = output_n.select(0, n);
|
||||
AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_forward_cuda", ([&] {
|
||||
ms_deformable_im2col_cuda(at::cuda::getCurrentCUDAStream(),
|
||||
value.data<scalar_t>() + n * im2col_step_ * per_value_size,
|
||||
spatial_shapes.data<int64_t>(),
|
||||
level_start_index.data<int64_t>(),
|
||||
sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
|
||||
attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
|
||||
batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
|
||||
columns.data<scalar_t>());
|
||||
|
||||
}));
|
||||
}
|
||||
|
||||
output = output.view({batch, num_query, num_heads*channels});
|
||||
|
||||
return output;
|
||||
}
|
||||
|
||||
|
||||
std::vector<at::Tensor> ms_deform_attn_cuda_backward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const at::Tensor &grad_output,
|
||||
const int im2col_step)
|
||||
{
|
||||
|
||||
AT_ASSERTM(value.is_contiguous(), "value tensor has to be contiguous");
|
||||
AT_ASSERTM(spatial_shapes.is_contiguous(), "spatial_shapes tensor has to be contiguous");
|
||||
AT_ASSERTM(level_start_index.is_contiguous(), "level_start_index tensor has to be contiguous");
|
||||
AT_ASSERTM(sampling_loc.is_contiguous(), "sampling_loc tensor has to be contiguous");
|
||||
AT_ASSERTM(attn_weight.is_contiguous(), "attn_weight tensor has to be contiguous");
|
||||
AT_ASSERTM(grad_output.is_contiguous(), "grad_output tensor has to be contiguous");
|
||||
|
||||
AT_ASSERTM(value.type().is_cuda(), "value must be a CUDA tensor");
|
||||
AT_ASSERTM(spatial_shapes.type().is_cuda(), "spatial_shapes must be a CUDA tensor");
|
||||
AT_ASSERTM(level_start_index.type().is_cuda(), "level_start_index must be a CUDA tensor");
|
||||
AT_ASSERTM(sampling_loc.type().is_cuda(), "sampling_loc must be a CUDA tensor");
|
||||
AT_ASSERTM(attn_weight.type().is_cuda(), "attn_weight must be a CUDA tensor");
|
||||
AT_ASSERTM(grad_output.type().is_cuda(), "grad_output must be a CUDA tensor");
|
||||
|
||||
const int batch = value.size(0);
|
||||
const int spatial_size = value.size(1);
|
||||
const int num_heads = value.size(2);
|
||||
const int channels = value.size(3);
|
||||
|
||||
const int num_levels = spatial_shapes.size(0);
|
||||
|
||||
const int num_query = sampling_loc.size(1);
|
||||
const int num_point = sampling_loc.size(4);
|
||||
|
||||
const int im2col_step_ = std::min(batch, im2col_step);
|
||||
|
||||
AT_ASSERTM(batch % im2col_step_ == 0, "batch(%d) must divide im2col_step(%d)", batch, im2col_step_);
|
||||
|
||||
auto grad_value = at::zeros_like(value);
|
||||
auto grad_sampling_loc = at::zeros_like(sampling_loc);
|
||||
auto grad_attn_weight = at::zeros_like(attn_weight);
|
||||
|
||||
const int batch_n = im2col_step_;
|
||||
auto per_value_size = spatial_size * num_heads * channels;
|
||||
auto per_sample_loc_size = num_query * num_heads * num_levels * num_point * 2;
|
||||
auto per_attn_weight_size = num_query * num_heads * num_levels * num_point;
|
||||
auto grad_output_n = grad_output.view({batch/im2col_step_, batch_n, num_query, num_heads, channels});
|
||||
|
||||
for (int n = 0; n < batch/im2col_step_; ++n)
|
||||
{
|
||||
auto grad_output_g = grad_output_n.select(0, n);
|
||||
AT_DISPATCH_FLOATING_TYPES(value.type(), "ms_deform_attn_backward_cuda", ([&] {
|
||||
ms_deformable_col2im_cuda(at::cuda::getCurrentCUDAStream(),
|
||||
grad_output_g.data<scalar_t>(),
|
||||
value.data<scalar_t>() + n * im2col_step_ * per_value_size,
|
||||
spatial_shapes.data<int64_t>(),
|
||||
level_start_index.data<int64_t>(),
|
||||
sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
|
||||
attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size,
|
||||
batch_n, spatial_size, num_heads, channels, num_levels, num_query, num_point,
|
||||
grad_value.data<scalar_t>() + n * im2col_step_ * per_value_size,
|
||||
grad_sampling_loc.data<scalar_t>() + n * im2col_step_ * per_sample_loc_size,
|
||||
grad_attn_weight.data<scalar_t>() + n * im2col_step_ * per_attn_weight_size);
|
||||
|
||||
}));
|
||||
}
|
||||
|
||||
return {
|
||||
grad_value, grad_sampling_loc, grad_attn_weight
|
||||
};
|
||||
}
|
||||
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,33 @@
|
|||
/*!
|
||||
**************************************************************************************************
|
||||
* Deformable DETR
|
||||
* Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
* Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
**************************************************************************************************
|
||||
* Modified from https://github.com/chengdazhi/Deformable-Convolution-V2-PyTorch/tree/pytorch_1.0.0
|
||||
**************************************************************************************************
|
||||
*/
|
||||
|
||||
#pragma once
|
||||
#include <torch/extension.h>
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
at::Tensor ms_deform_attn_cuda_forward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const int im2col_step);
|
||||
|
||||
std::vector<at::Tensor> ms_deform_attn_cuda_backward(
|
||||
const at::Tensor &value,
|
||||
const at::Tensor &spatial_shapes,
|
||||
const at::Tensor &level_start_index,
|
||||
const at::Tensor &sampling_loc,
|
||||
const at::Tensor &attn_weight,
|
||||
const at::Tensor &grad_output,
|
||||
const int im2col_step);
|
||||
|
||||
} // namespace groundingdino
|
||||
File diff suppressed because it is too large
Load Diff
|
|
@ -0,0 +1,7 @@
|
|||
#include <cuda_runtime_api.h>
|
||||
|
||||
namespace groundingdino {
|
||||
int get_cudart_version() {
|
||||
return CUDART_VERSION;
|
||||
}
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,58 @@
|
|||
// Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
|
||||
#include "MsDeformAttn/ms_deform_attn.h"
|
||||
|
||||
namespace groundingdino {
|
||||
|
||||
#ifdef WITH_CUDA
|
||||
extern int get_cudart_version();
|
||||
#endif
|
||||
|
||||
std::string get_cuda_version() {
|
||||
#ifdef WITH_CUDA
|
||||
std::ostringstream oss;
|
||||
|
||||
// copied from
|
||||
// https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/cuda/detail/CUDAHooks.cpp#L231
|
||||
auto printCudaStyleVersion = [&](int v) {
|
||||
oss << (v / 1000) << "." << (v / 10 % 100);
|
||||
if (v % 10 != 0) {
|
||||
oss << "." << (v % 10);
|
||||
}
|
||||
};
|
||||
printCudaStyleVersion(get_cudart_version());
|
||||
return oss.str();
|
||||
#else
|
||||
return std::string("not available");
|
||||
#endif
|
||||
}
|
||||
|
||||
// similar to
|
||||
// https://github.com/pytorch/pytorch/blob/master/aten/src/ATen/Version.cpp
|
||||
std::string get_compiler_version() {
|
||||
std::ostringstream ss;
|
||||
#if defined(__GNUC__)
|
||||
#ifndef __clang__
|
||||
{ ss << "GCC " << __GNUC__ << "." << __GNUC_MINOR__; }
|
||||
#endif
|
||||
#endif
|
||||
|
||||
#if defined(__clang_major__)
|
||||
{
|
||||
ss << "clang " << __clang_major__ << "." << __clang_minor__ << "."
|
||||
<< __clang_patchlevel__;
|
||||
}
|
||||
#endif
|
||||
|
||||
#if defined(_MSC_VER)
|
||||
{ ss << "MSVC " << _MSC_FULL_VER; }
|
||||
#endif
|
||||
return ss.str();
|
||||
}
|
||||
|
||||
PYBIND11_MODULE(TORCH_EXTENSION_NAME, m) {
|
||||
m.def("ms_deform_attn_forward", &ms_deform_attn_forward, "ms_deform_attn_forward");
|
||||
m.def("ms_deform_attn_backward", &ms_deform_attn_backward, "ms_deform_attn_backward");
|
||||
}
|
||||
|
||||
} // namespace groundingdino
|
||||
|
|
@ -0,0 +1,297 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from timm.models.layers import DropPath
|
||||
|
||||
|
||||
class FeatureResizer(nn.Module):
|
||||
"""
|
||||
This class takes as input a set of embeddings of dimension C1 and outputs a set of
|
||||
embedding of dimension C2, after a linear transformation, dropout and normalization (LN).
|
||||
"""
|
||||
|
||||
def __init__(self, input_feat_size, output_feat_size, dropout, do_ln=True):
|
||||
super().__init__()
|
||||
self.do_ln = do_ln
|
||||
# Object feature encoding
|
||||
self.fc = nn.Linear(input_feat_size, output_feat_size, bias=True)
|
||||
self.layer_norm = nn.LayerNorm(output_feat_size, eps=1e-12)
|
||||
self.dropout = nn.Dropout(dropout)
|
||||
|
||||
def forward(self, encoder_features):
|
||||
x = self.fc(encoder_features)
|
||||
if self.do_ln:
|
||||
x = self.layer_norm(x)
|
||||
output = self.dropout(x)
|
||||
return output
|
||||
|
||||
|
||||
def l1norm(X, dim, eps=1e-8):
|
||||
"""L1-normalize columns of X"""
|
||||
norm = torch.abs(X).sum(dim=dim, keepdim=True) + eps
|
||||
X = torch.div(X, norm)
|
||||
return X
|
||||
|
||||
|
||||
def l2norm(X, dim, eps=1e-8):
|
||||
"""L2-normalize columns of X"""
|
||||
norm = torch.pow(X, 2).sum(dim=dim, keepdim=True).sqrt() + eps
|
||||
X = torch.div(X, norm)
|
||||
return X
|
||||
|
||||
|
||||
def func_attention(query, context, smooth=1, raw_feature_norm="softmax", eps=1e-8):
|
||||
"""
|
||||
query: (n_context, queryL, d)
|
||||
context: (n_context, sourceL, d)
|
||||
"""
|
||||
batch_size_q, queryL = query.size(0), query.size(1)
|
||||
batch_size, sourceL = context.size(0), context.size(1)
|
||||
|
||||
# Get attention
|
||||
# --> (batch, d, queryL)
|
||||
queryT = torch.transpose(query, 1, 2)
|
||||
|
||||
# (batch, sourceL, d)(batch, d, queryL)
|
||||
# --> (batch, sourceL, queryL)
|
||||
attn = torch.bmm(context, queryT)
|
||||
if raw_feature_norm == "softmax":
|
||||
# --> (batch*sourceL, queryL)
|
||||
attn = attn.view(batch_size * sourceL, queryL)
|
||||
attn = nn.Softmax()(attn)
|
||||
# --> (batch, sourceL, queryL)
|
||||
attn = attn.view(batch_size, sourceL, queryL)
|
||||
elif raw_feature_norm == "l2norm":
|
||||
attn = l2norm(attn, 2)
|
||||
elif raw_feature_norm == "clipped_l2norm":
|
||||
attn = nn.LeakyReLU(0.1)(attn)
|
||||
attn = l2norm(attn, 2)
|
||||
else:
|
||||
raise ValueError("unknown first norm type:", raw_feature_norm)
|
||||
# --> (batch, queryL, sourceL)
|
||||
attn = torch.transpose(attn, 1, 2).contiguous()
|
||||
# --> (batch*queryL, sourceL)
|
||||
attn = attn.view(batch_size * queryL, sourceL)
|
||||
attn = nn.Softmax()(attn * smooth)
|
||||
# --> (batch, queryL, sourceL)
|
||||
attn = attn.view(batch_size, queryL, sourceL)
|
||||
# --> (batch, sourceL, queryL)
|
||||
attnT = torch.transpose(attn, 1, 2).contiguous()
|
||||
|
||||
# --> (batch, d, sourceL)
|
||||
contextT = torch.transpose(context, 1, 2)
|
||||
# (batch x d x sourceL)(batch x sourceL x queryL)
|
||||
# --> (batch, d, queryL)
|
||||
weightedContext = torch.bmm(contextT, attnT)
|
||||
# --> (batch, queryL, d)
|
||||
weightedContext = torch.transpose(weightedContext, 1, 2)
|
||||
|
||||
return weightedContext, attnT
|
||||
|
||||
|
||||
class BiMultiHeadAttention(nn.Module):
|
||||
def __init__(self, v_dim, l_dim, embed_dim, num_heads, dropout=0.1, cfg=None):
|
||||
super(BiMultiHeadAttention, self).__init__()
|
||||
|
||||
self.embed_dim = embed_dim
|
||||
self.num_heads = num_heads
|
||||
self.head_dim = embed_dim // num_heads
|
||||
self.v_dim = v_dim
|
||||
self.l_dim = l_dim
|
||||
|
||||
assert (
|
||||
self.head_dim * self.num_heads == self.embed_dim
|
||||
), f"embed_dim must be divisible by num_heads (got `embed_dim`: {self.embed_dim} and `num_heads`: {self.num_heads})."
|
||||
self.scale = self.head_dim ** (-0.5)
|
||||
self.dropout = dropout
|
||||
|
||||
self.v_proj = nn.Linear(self.v_dim, self.embed_dim)
|
||||
self.l_proj = nn.Linear(self.l_dim, self.embed_dim)
|
||||
self.values_v_proj = nn.Linear(self.v_dim, self.embed_dim)
|
||||
self.values_l_proj = nn.Linear(self.l_dim, self.embed_dim)
|
||||
|
||||
self.out_v_proj = nn.Linear(self.embed_dim, self.v_dim)
|
||||
self.out_l_proj = nn.Linear(self.embed_dim, self.l_dim)
|
||||
|
||||
self.stable_softmax_2d = True
|
||||
self.clamp_min_for_underflow = True
|
||||
self.clamp_max_for_overflow = True
|
||||
|
||||
self._reset_parameters()
|
||||
|
||||
def _shape(self, tensor: torch.Tensor, seq_len: int, bsz: int):
|
||||
return tensor.view(bsz, seq_len, self.num_heads, self.head_dim).transpose(1, 2).contiguous()
|
||||
|
||||
def _reset_parameters(self):
|
||||
nn.init.xavier_uniform_(self.v_proj.weight)
|
||||
self.v_proj.bias.data.fill_(0)
|
||||
nn.init.xavier_uniform_(self.l_proj.weight)
|
||||
self.l_proj.bias.data.fill_(0)
|
||||
nn.init.xavier_uniform_(self.values_v_proj.weight)
|
||||
self.values_v_proj.bias.data.fill_(0)
|
||||
nn.init.xavier_uniform_(self.values_l_proj.weight)
|
||||
self.values_l_proj.bias.data.fill_(0)
|
||||
nn.init.xavier_uniform_(self.out_v_proj.weight)
|
||||
self.out_v_proj.bias.data.fill_(0)
|
||||
nn.init.xavier_uniform_(self.out_l_proj.weight)
|
||||
self.out_l_proj.bias.data.fill_(0)
|
||||
|
||||
def forward(self, v, l, attention_mask_v=None, attention_mask_l=None):
|
||||
"""_summary_
|
||||
|
||||
Args:
|
||||
v (_type_): bs, n_img, dim
|
||||
l (_type_): bs, n_text, dim
|
||||
attention_mask_v (_type_, optional): _description_. bs, n_img
|
||||
attention_mask_l (_type_, optional): _description_. bs, n_text
|
||||
|
||||
Returns:
|
||||
_type_: _description_
|
||||
"""
|
||||
# if os.environ.get('IPDB_SHILONG_DEBUG', None) == 'INFO':
|
||||
# import ipdb; ipdb.set_trace()
|
||||
bsz, tgt_len, _ = v.size()
|
||||
|
||||
query_states = self.v_proj(v) * self.scale
|
||||
key_states = self._shape(self.l_proj(l), -1, bsz)
|
||||
value_v_states = self._shape(self.values_v_proj(v), -1, bsz)
|
||||
value_l_states = self._shape(self.values_l_proj(l), -1, bsz)
|
||||
|
||||
proj_shape = (bsz * self.num_heads, -1, self.head_dim)
|
||||
query_states = self._shape(query_states, tgt_len, bsz).view(*proj_shape)
|
||||
key_states = key_states.view(*proj_shape)
|
||||
value_v_states = value_v_states.view(*proj_shape)
|
||||
value_l_states = value_l_states.view(*proj_shape)
|
||||
|
||||
src_len = key_states.size(1)
|
||||
attn_weights = torch.bmm(query_states, key_states.transpose(1, 2)) # bs*nhead, nimg, ntxt
|
||||
|
||||
if attn_weights.size() != (bsz * self.num_heads, tgt_len, src_len):
|
||||
raise ValueError(
|
||||
f"Attention weights should be of size {(bsz * self.num_heads, tgt_len, src_len)}, but is {attn_weights.size()}"
|
||||
)
|
||||
|
||||
if self.stable_softmax_2d:
|
||||
attn_weights = attn_weights - attn_weights.max()
|
||||
|
||||
if self.clamp_min_for_underflow:
|
||||
attn_weights = torch.clamp(
|
||||
attn_weights, min=-50000
|
||||
) # Do not increase -50000, data type half has quite limited range
|
||||
if self.clamp_max_for_overflow:
|
||||
attn_weights = torch.clamp(
|
||||
attn_weights, max=50000
|
||||
) # Do not increase 50000, data type half has quite limited range
|
||||
|
||||
attn_weights_T = attn_weights.transpose(1, 2)
|
||||
attn_weights_l = attn_weights_T - torch.max(attn_weights_T, dim=-1, keepdim=True)[0]
|
||||
if self.clamp_min_for_underflow:
|
||||
attn_weights_l = torch.clamp(
|
||||
attn_weights_l, min=-50000
|
||||
) # Do not increase -50000, data type half has quite limited range
|
||||
if self.clamp_max_for_overflow:
|
||||
attn_weights_l = torch.clamp(
|
||||
attn_weights_l, max=50000
|
||||
) # Do not increase 50000, data type half has quite limited range
|
||||
|
||||
# mask vison for language
|
||||
if attention_mask_v is not None:
|
||||
attention_mask_v = (
|
||||
attention_mask_v[:, None, None, :].repeat(1, self.num_heads, 1, 1).flatten(0, 1)
|
||||
)
|
||||
attn_weights_l.masked_fill_(attention_mask_v, float("-inf"))
|
||||
|
||||
attn_weights_l = attn_weights_l.softmax(dim=-1)
|
||||
|
||||
# mask language for vision
|
||||
if attention_mask_l is not None:
|
||||
attention_mask_l = (
|
||||
attention_mask_l[:, None, None, :].repeat(1, self.num_heads, 1, 1).flatten(0, 1)
|
||||
)
|
||||
attn_weights.masked_fill_(attention_mask_l, float("-inf"))
|
||||
attn_weights_v = attn_weights.softmax(dim=-1)
|
||||
|
||||
attn_probs_v = F.dropout(attn_weights_v, p=self.dropout, training=self.training)
|
||||
attn_probs_l = F.dropout(attn_weights_l, p=self.dropout, training=self.training)
|
||||
|
||||
attn_output_v = torch.bmm(attn_probs_v, value_l_states)
|
||||
attn_output_l = torch.bmm(attn_probs_l, value_v_states)
|
||||
|
||||
if attn_output_v.size() != (bsz * self.num_heads, tgt_len, self.head_dim):
|
||||
raise ValueError(
|
||||
f"`attn_output_v` should be of size {(bsz, self.num_heads, tgt_len, self.head_dim)}, but is {attn_output_v.size()}"
|
||||
)
|
||||
|
||||
if attn_output_l.size() != (bsz * self.num_heads, src_len, self.head_dim):
|
||||
raise ValueError(
|
||||
f"`attn_output_l` should be of size {(bsz, self.num_heads, src_len, self.head_dim)}, but is {attn_output_l.size()}"
|
||||
)
|
||||
|
||||
attn_output_v = attn_output_v.view(bsz, self.num_heads, tgt_len, self.head_dim)
|
||||
attn_output_v = attn_output_v.transpose(1, 2)
|
||||
attn_output_v = attn_output_v.reshape(bsz, tgt_len, self.embed_dim)
|
||||
|
||||
attn_output_l = attn_output_l.view(bsz, self.num_heads, src_len, self.head_dim)
|
||||
attn_output_l = attn_output_l.transpose(1, 2)
|
||||
attn_output_l = attn_output_l.reshape(bsz, src_len, self.embed_dim)
|
||||
|
||||
attn_output_v = self.out_v_proj(attn_output_v)
|
||||
attn_output_l = self.out_l_proj(attn_output_l)
|
||||
|
||||
return attn_output_v, attn_output_l
|
||||
|
||||
|
||||
# Bi-Direction MHA (text->image, image->text)
|
||||
class BiAttentionBlock(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
v_dim,
|
||||
l_dim,
|
||||
embed_dim,
|
||||
num_heads,
|
||||
dropout=0.1,
|
||||
drop_path=0.0,
|
||||
init_values=1e-4,
|
||||
cfg=None,
|
||||
):
|
||||
"""
|
||||
Inputs:
|
||||
embed_dim - Dimensionality of input and attention feature vectors
|
||||
hidden_dim - Dimensionality of hidden layer in feed-forward network
|
||||
(usually 2-4x larger than embed_dim)
|
||||
num_heads - Number of heads to use in the Multi-Head Attention block
|
||||
dropout - Amount of dropout to apply in the feed-forward network
|
||||
"""
|
||||
super(BiAttentionBlock, self).__init__()
|
||||
|
||||
# pre layer norm
|
||||
self.layer_norm_v = nn.LayerNorm(v_dim)
|
||||
self.layer_norm_l = nn.LayerNorm(l_dim)
|
||||
self.attn = BiMultiHeadAttention(
|
||||
v_dim=v_dim, l_dim=l_dim, embed_dim=embed_dim, num_heads=num_heads, dropout=dropout
|
||||
)
|
||||
|
||||
# add layer scale for training stability
|
||||
self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity()
|
||||
self.gamma_v = nn.Parameter(init_values * torch.ones((v_dim)), requires_grad=True)
|
||||
self.gamma_l = nn.Parameter(init_values * torch.ones((l_dim)), requires_grad=True)
|
||||
|
||||
def forward(self, v, l, attention_mask_v=None, attention_mask_l=None):
|
||||
v = self.layer_norm_v(v)
|
||||
l = self.layer_norm_l(l)
|
||||
delta_v, delta_l = self.attn(
|
||||
v, l, attention_mask_v=attention_mask_v, attention_mask_l=attention_mask_l
|
||||
)
|
||||
# v, l = v + delta_v, l + delta_l
|
||||
v = v + self.drop_path(self.gamma_v * delta_v)
|
||||
l = l + self.drop_path(self.gamma_l * delta_l)
|
||||
return v, l
|
||||
|
||||
# def forward(self, v:List[torch.Tensor], l, attention_mask_v=None, attention_mask_l=None)
|
||||
|
|
@ -0,0 +1,708 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Conditional DETR model and criterion classes.
|
||||
# Copyright (c) 2021 Microsoft. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Modified from DETR (https://github.com/facebookresearch/detr)
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
# Modified from Deformable DETR (https://github.com/fundamentalvision/Deformable-DETR)
|
||||
# Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
import copy
|
||||
from typing import List
|
||||
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import nn, einsum
|
||||
from torchvision.ops.boxes import nms
|
||||
from transformers import AutoTokenizer, BertModel, BertTokenizer, RobertaModel, RobertaTokenizerFast
|
||||
|
||||
from groundingdino_new.util import box_ops, get_tokenlizer
|
||||
from groundingdino_new.util.misc import (
|
||||
NestedTensor,
|
||||
accuracy,
|
||||
get_world_size,
|
||||
interpolate,
|
||||
inverse_sigmoid,
|
||||
is_dist_avail_and_initialized,
|
||||
nested_tensor_from_tensor_list,
|
||||
)
|
||||
from groundingdino_new.util.utils import get_phrases_from_posmap
|
||||
from groundingdino_new.util.visualizer import COCOVisualizer
|
||||
from groundingdino_new.util.vl_utils import create_positive_map_from_span
|
||||
|
||||
from ..registry import MODULE_BUILD_FUNCS
|
||||
from .backbone import build_backbone
|
||||
from .bertwarper import (
|
||||
BertModelWarper,
|
||||
generate_masks_with_special_tokens,
|
||||
generate_masks_with_special_tokens_and_transfer_map,
|
||||
)
|
||||
from .transformer import build_transformer
|
||||
from .utils import MLP, ContrastiveEmbed, sigmoid_focal_loss
|
||||
from maskrcnn_benchmark.structures.image_list import ImageList
|
||||
from maskrcnn_benchmark.modeling.rpn.inference import convert_grounding_to_od_logits
|
||||
from maskrcnn_benchmark.modeling.box_coder import BoxCoder
|
||||
from maskrcnn_benchmark.structures.bounding_box import BoxList
|
||||
from maskrcnn_benchmark.structures.boxlist_ops import remove_small_boxes
|
||||
from maskrcnn_benchmark.structures.boxlist_ops import boxlist_ml_nms
|
||||
from maskrcnn_benchmark.structures.boxlist_ops import cat_boxlist
|
||||
# from groundingdino_new.util.inference import preprocess_caption
|
||||
from maskrcnn_benchmark.modeling.poolers import CustomPooler, Pooler
|
||||
from groundingdino_new.models.GroundingDINO.loss import SetCriterion
|
||||
from groundingdino_new.models.GroundingDINO.matcher import build_matcher
|
||||
from maskrcnn_benchmark.modeling.language_backbone import build_language_backbone
|
||||
from maskrcnn_benchmark.modeling.language_backbone.modeling_bert_new import QVBertModel
|
||||
from transformers import BertConfig, RobertaConfig, RobertaModel
|
||||
from maskrcnn_benchmark.modeling.query_selector import build_query_selector
|
||||
|
||||
import os
|
||||
|
||||
def expand_bbox(box_list, expand_ratio=1.5):
|
||||
new_box_list=[]
|
||||
for boxes in box_list:
|
||||
assert boxes.mode == "xyxy"
|
||||
bbox=boxes.bbox
|
||||
image_size=boxes.size
|
||||
box_w, box_h = bbox[:,2] - bbox[:,0], bbox[:,3] - bbox[:,1]
|
||||
new_box_w, new_box_h = box_w*expand_ratio, box_h*expand_ratio
|
||||
diff_w=(new_box_w-box_w)/2
|
||||
diff_h=(new_box_h-box_h)/2
|
||||
diff=torch.stack([-diff_w, -diff_h, diff_w, diff_h], dim=1)
|
||||
new_bbox=bbox+diff
|
||||
new_boxes=BoxList(new_bbox, image_size, mode="xyxy")
|
||||
labels=boxes.get_field('labels')
|
||||
new_boxes.add_field('labels', labels)
|
||||
new_boxes=new_boxes.clip_to_image(remove_empty=True)
|
||||
new_box_list.append(new_boxes)
|
||||
return new_box_list
|
||||
|
||||
def preprocess_caption(caption: str) -> str:
|
||||
result = caption.lower().strip()
|
||||
if result.endswith("."):
|
||||
return result
|
||||
return result + "."
|
||||
|
||||
class GroundingDINO(nn.Module):
|
||||
"""This is the Cross-Attention Detector module that performs object detection"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
backbone,
|
||||
transformer,
|
||||
num_queries,
|
||||
aux_loss=False,
|
||||
iter_update=False,
|
||||
query_dim=2,
|
||||
num_feature_levels=1,
|
||||
nheads=8,
|
||||
# two stage
|
||||
two_stage_type="no", # ['no', 'standard']
|
||||
dec_pred_bbox_embed_share=True,
|
||||
two_stage_class_embed_share=True,
|
||||
two_stage_bbox_embed_share=True,
|
||||
num_patterns=0,
|
||||
dn_number=100,
|
||||
dn_box_noise_scale=0.4,
|
||||
dn_label_noise_ratio=0.5,
|
||||
dn_labelbook_size=100,
|
||||
text_encoder_type="bert-base-uncased",
|
||||
sub_sentence_present=True,
|
||||
max_text_len=256,
|
||||
cfg = None,
|
||||
):
|
||||
"""Initializes the model.
|
||||
Parameters:
|
||||
backbone: torch module of the backbone to be used. See backbone.py
|
||||
transformer: torch module of the transformer architecture. See transformer.py
|
||||
num_queries: number of object queries, ie detection slot. This is the maximal number of objects
|
||||
Conditional DETR can detect in a single image. For COCO, we recommend 100 queries.
|
||||
aux_loss: True if auxiliary decoding losses (loss at each decoder layer) are to be used.
|
||||
"""
|
||||
super().__init__()
|
||||
self.cfg = cfg
|
||||
self.box_threshold = cfg.GROUNDINGDINO.box_threshold
|
||||
self.num_queries = num_queries
|
||||
self.transformer = transformer
|
||||
self.hidden_dim = hidden_dim = transformer.d_model
|
||||
self.num_feature_levels = num_feature_levels
|
||||
self.nheads = nheads
|
||||
self.max_text_len = 256
|
||||
self.sub_sentence_present = sub_sentence_present
|
||||
|
||||
# setting query dim
|
||||
self.query_dim = query_dim
|
||||
assert query_dim == 4
|
||||
|
||||
# for dn training
|
||||
self.num_patterns = num_patterns
|
||||
self.dn_number = dn_number
|
||||
self.dn_box_noise_scale = dn_box_noise_scale
|
||||
self.dn_label_noise_ratio = dn_label_noise_ratio
|
||||
self.dn_labelbook_size = dn_labelbook_size
|
||||
|
||||
# loss criterion
|
||||
self.loss_evaluator = SetCriterion(matcher=build_matcher(cfg.GROUNDINGDINO.matcher), cfg=cfg)
|
||||
|
||||
|
||||
# box pooler for extracting cache
|
||||
resolution = cfg.MODEL.ROI_BOX_HEAD.POOLER_RESOLUTION
|
||||
if cfg.VISION_QUERY.SELECT_FPN_LEVEL:
|
||||
self.pooler = Pooler(
|
||||
output_size= (resolution, resolution) ,
|
||||
scales=cfg.MODEL.ROI_BOX_HEAD.POOLER_SCALES,
|
||||
sampling_ratio=cfg.MODEL.ROI_BOX_HEAD.POOLER_SAMPLING_RATIO,
|
||||
use_v2=True,
|
||||
)
|
||||
else:
|
||||
self.pooler = CustomPooler(
|
||||
output_size= (resolution, resolution) ,
|
||||
scales=cfg.MODEL.ROI_BOX_HEAD.POOLER_SCALES,
|
||||
sampling_ratio=cfg.MODEL.ROI_BOX_HEAD.POOLER_SAMPLING_RATIO,
|
||||
use_v2=True,
|
||||
)
|
||||
self.pool=nn.AvgPool2d(2)
|
||||
|
||||
# query selector
|
||||
if cfg.VISION_QUERY.DISABLE_SELECTOR:
|
||||
self.query_selector = None
|
||||
else:
|
||||
self.query_selector = build_query_selector(cfg)
|
||||
|
||||
# bert
|
||||
self.tokenizer = get_tokenlizer.get_tokenlizer(text_encoder_type)
|
||||
if os.path.basename(text_encoder_type) != "bert-base-uncased":
|
||||
raise NotImplementedError
|
||||
# self.bert = get_tokenlizer.get_pretrained_language_model(text_encoder_type)
|
||||
config = BertConfig.from_pretrained(text_encoder_type)
|
||||
self.bert = QVBertModel.from_pretrained(text_encoder_type, dim_t=config.hidden_size, dim_v=self.hidden_dim, share_kv=cfg.VISION_QUERY.SHARE_KV, cfg=cfg, config=config)
|
||||
|
||||
self.bert.pooler.dense.weight.requires_grad_(False)
|
||||
self.bert.pooler.dense.bias.requires_grad_(False)
|
||||
self.bert = BertModelWarper(bert_model=self.bert)
|
||||
|
||||
self.feat_map = nn.Linear(self.bert.config.hidden_size, self.hidden_dim, bias=True)
|
||||
nn.init.constant_(self.feat_map.bias.data, 0)
|
||||
nn.init.xavier_uniform_(self.feat_map.weight.data)
|
||||
# freeze
|
||||
|
||||
# special tokens
|
||||
self.specical_tokens = self.tokenizer.convert_tokens_to_ids(["[CLS]", "[SEP]", ".", "?"])
|
||||
|
||||
# prepare input projection layers
|
||||
if num_feature_levels > 1:
|
||||
num_backbone_outs = len(backbone.num_channels)
|
||||
input_proj_list = []
|
||||
for _ in range(num_backbone_outs):
|
||||
in_channels = backbone.num_channels[_]
|
||||
input_proj_list.append(
|
||||
nn.Sequential(
|
||||
nn.Conv2d(in_channels, hidden_dim, kernel_size=1),
|
||||
nn.GroupNorm(32, hidden_dim),
|
||||
)
|
||||
)
|
||||
for _ in range(num_feature_levels - num_backbone_outs):
|
||||
input_proj_list.append(
|
||||
nn.Sequential(
|
||||
nn.Conv2d(in_channels, hidden_dim, kernel_size=3, stride=2, padding=1),
|
||||
nn.GroupNorm(32, hidden_dim),
|
||||
)
|
||||
)
|
||||
in_channels = hidden_dim
|
||||
self.input_proj = nn.ModuleList(input_proj_list)
|
||||
else:
|
||||
assert two_stage_type == "no", "two_stage_type should be no if num_feature_levels=1 !!!"
|
||||
self.input_proj = nn.ModuleList(
|
||||
[
|
||||
nn.Sequential(
|
||||
nn.Conv2d(backbone.num_channels[-1], hidden_dim, kernel_size=1),
|
||||
nn.GroupNorm(32, hidden_dim),
|
||||
)
|
||||
]
|
||||
)
|
||||
|
||||
self.backbone = backbone
|
||||
self.aux_loss = aux_loss
|
||||
self.box_pred_damping = box_pred_damping = None
|
||||
|
||||
self.iter_update = iter_update
|
||||
assert iter_update, "Why not iter_update?"
|
||||
|
||||
# prepare pred layers
|
||||
self.dec_pred_bbox_embed_share = dec_pred_bbox_embed_share
|
||||
# prepare class & box embed
|
||||
_class_embed = ContrastiveEmbed()
|
||||
|
||||
_bbox_embed = MLP(hidden_dim, hidden_dim, 4, 3)
|
||||
nn.init.constant_(_bbox_embed.layers[-1].weight.data, 0)
|
||||
nn.init.constant_(_bbox_embed.layers[-1].bias.data, 0)
|
||||
|
||||
if dec_pred_bbox_embed_share:
|
||||
box_embed_layerlist = [_bbox_embed for i in range(transformer.num_decoder_layers)]
|
||||
else:
|
||||
box_embed_layerlist = [
|
||||
copy.deepcopy(_bbox_embed) for i in range(transformer.num_decoder_layers)
|
||||
]
|
||||
class_embed_layerlist = [_class_embed for i in range(transformer.num_decoder_layers)]
|
||||
self.bbox_embed = nn.ModuleList(box_embed_layerlist)
|
||||
self.class_embed = nn.ModuleList(class_embed_layerlist)
|
||||
self.transformer.decoder.bbox_embed = self.bbox_embed
|
||||
self.transformer.decoder.class_embed = self.class_embed
|
||||
|
||||
# two stage
|
||||
self.two_stage_type = two_stage_type
|
||||
assert two_stage_type in ["no", "standard"], "unknown param {} of two_stage_type".format(
|
||||
two_stage_type
|
||||
)
|
||||
if two_stage_type != "no":
|
||||
if two_stage_bbox_embed_share:
|
||||
assert dec_pred_bbox_embed_share
|
||||
self.transformer.enc_out_bbox_embed = _bbox_embed
|
||||
else:
|
||||
self.transformer.enc_out_bbox_embed = copy.deepcopy(_bbox_embed)
|
||||
|
||||
if two_stage_class_embed_share:
|
||||
assert dec_pred_bbox_embed_share
|
||||
self.transformer.enc_out_class_embed = _class_embed
|
||||
else:
|
||||
self.transformer.enc_out_class_embed = copy.deepcopy(_class_embed)
|
||||
|
||||
self.refpoint_embed = None
|
||||
|
||||
self._reset_parameters()
|
||||
|
||||
def _reset_parameters(self):
|
||||
# init input_proj
|
||||
for proj in self.input_proj:
|
||||
nn.init.xavier_uniform_(proj[0].weight, gain=1)
|
||||
nn.init.constant_(proj[0].bias, 0)
|
||||
|
||||
def init_ref_points(self, use_num_queries):
|
||||
self.refpoint_embed = nn.Embedding(use_num_queries, self.query_dim)
|
||||
|
||||
|
||||
def convert_groundingdino_to_glip_output(self, groundingdino_out, positive_map, image_sizes):
|
||||
dot_product_logits = groundingdino_out['pred_logits']
|
||||
box_regression = groundingdino_out['pred_boxes']
|
||||
B, N, _ = dot_product_logits.shape
|
||||
box_cls = dot_product_logits.new_zeros(B, N, self.cfg.MODEL.DYHEAD.NUM_CLASSES - 1)
|
||||
# candidate_inds = dot_product_logits.max(dim=-1)[0] > self.box_threshold
|
||||
scores = convert_grounding_to_od_logits(logits=dot_product_logits, box_cls=box_cls,
|
||||
positive_map=positive_map,
|
||||
score_agg="MEAN",
|
||||
)
|
||||
box_cls = scores
|
||||
|
||||
candidate_inds = box_cls.max(dim=-1)[0] > self.box_threshold
|
||||
# pre_nms_top_n = candidate_inds.reshape(N, -1).sum(1)
|
||||
# pre_nms_top_n = pre_nms_top_n.clamp(max=self.pre_nms_top_n)
|
||||
|
||||
results = []
|
||||
for per_box_cls, per_box_regression, per_candidate_inds, image_size \
|
||||
in zip(box_cls, box_regression, candidate_inds, image_sizes):
|
||||
per_box_cls = per_box_cls[per_candidate_inds]
|
||||
|
||||
per_box_cls, top_k_indices = per_box_cls.topk(1, sorted=False)
|
||||
|
||||
per_class = top_k_indices[:, 0] + 1
|
||||
|
||||
# print(per_class)
|
||||
|
||||
box = per_box_regression[per_candidate_inds, :].view(-1, 4)
|
||||
H, W = image_size
|
||||
# from 0..1 to 0..W, 0..H
|
||||
box = box * torch.Tensor([W, H, W, H]).to(box.device)[None, ...]
|
||||
# from xywh to xyxy
|
||||
box[:, :2] = box[:, :2] - box[:, 2:] / 2
|
||||
box[:, 2:] = box[:, 2:] + box[:, :2]
|
||||
|
||||
detections = box
|
||||
|
||||
boxlist = BoxList(detections, (W, H), mode="xyxy")
|
||||
boxlist.add_field("labels", per_class)
|
||||
boxlist.add_field("scores", per_box_cls[:,0])
|
||||
boxlist = boxlist.clip_to_image(remove_empty=False)
|
||||
boxlist = remove_small_boxes(boxlist, min_size=0)
|
||||
results.append(boxlist)
|
||||
|
||||
return results
|
||||
|
||||
def load_query_bank(self, query_path):
|
||||
self.query_selector.load_query_bank(query_path)
|
||||
|
||||
@torch.no_grad()
|
||||
def extract_query(self,
|
||||
samples=None,
|
||||
targets=None,
|
||||
query_images=None, # default_dict(list) ,list[tensors] num_classes: (num_queries, num_scales, num_channels)
|
||||
visual_features=None,
|
||||
exclude_similar=False,
|
||||
device = None,
|
||||
max_query_number = None,
|
||||
):
|
||||
device = device if device else samples.tensors.device
|
||||
targets = [target.to(device)
|
||||
for target in targets if target is not None]
|
||||
targets=expand_bbox(targets, expand_ratio=self.cfg.VISION_QUERY.EXPAND_RATIO)
|
||||
if visual_features is None:
|
||||
if isinstance(samples, ImageList):
|
||||
image_sizes = samples.image_sizes
|
||||
samples = samples.tensors
|
||||
if isinstance(samples, (list, torch.Tensor)):
|
||||
samples = nested_tensor_from_tensor_list(samples, image_sizes=image_sizes)
|
||||
features, poss = self.backbone(samples)
|
||||
|
||||
srcs = []
|
||||
masks = []
|
||||
for l, feat in enumerate(features):
|
||||
src, mask = feat.decompose()
|
||||
srcs.append(self.input_proj[l](src))
|
||||
masks.append(mask)
|
||||
assert mask is not None
|
||||
if self.num_feature_levels > len(srcs):
|
||||
_len_srcs = len(srcs)
|
||||
for l in range(_len_srcs, self.num_feature_levels):
|
||||
if l == _len_srcs:
|
||||
src = self.input_proj[l](features[-1].tensors)
|
||||
else:
|
||||
src = self.input_proj[l](srcs[-1])
|
||||
m = samples.mask
|
||||
mask = F.interpolate(m[None].float(), size=src.shape[-2:]).to(torch.bool)[0]
|
||||
pos_l = self.backbone[1](NestedTensor(src, mask)).to(src.dtype)
|
||||
srcs.append(src)
|
||||
masks.append(mask)
|
||||
poss.append(pos_l)
|
||||
|
||||
visual_features = srcs
|
||||
else:
|
||||
visual_features = [v.to(device) for v in visual_features]
|
||||
|
||||
if self.cfg.VISION_QUERY.SELECT_FPN_LEVEL:
|
||||
query_feats=self.pooler(visual_features, targets) # num_boxes, num_channels, pooler_size, pooler_size
|
||||
query_feats=query_feats[None, ...] # 1, num_boxes, num_channels, pooler_size, pooler_size
|
||||
else:
|
||||
query_feats=self.pooler(visual_features, targets) # num_scales, num_boxes, num_channels, pooler_size, pooler_size
|
||||
|
||||
# average different fpn levels
|
||||
if not self.cfg.VISION_QUERY.SELECT_FPN_LEVEL:
|
||||
assert len(visual_features) == len(query_feats) == 5 # TODO: support flexible level numbers
|
||||
query_feats = query_feats.mean(dim=[-2,-1]).permute(1, 0, 2) # num_boxes, num_scales, num_channels
|
||||
|
||||
labels=torch.cat([t.get_field('labels') for t in targets])
|
||||
assert len(labels)==len(query_feats)
|
||||
|
||||
max_query_number = self.cfg.VISION_QUERY.MAX_QUERY_NUMBER if max_query_number is None else max_query_number
|
||||
for label, feat in zip(labels, query_feats):
|
||||
label=label.item()
|
||||
num_queries=len(query_images[label])
|
||||
if num_queries >= max_query_number:
|
||||
continue
|
||||
if exclude_similar and num_queries > 0:
|
||||
assert feat.shape[0] == 1 # TODO: enable all-level and spacial features
|
||||
bank_features = F.normalize(query_images[label], p=2, dim=-1) # N, 1, C
|
||||
new_features = F.normalize(feat, p=2, dim=-1) # 1, C
|
||||
similarity = einsum('b n d, n d -> b n', bank_features, new_features)
|
||||
has_similar_in_bank = (similarity > self.cfg.VISION_QUERY.SIMILARITY_THRESHOLD).sum() > 0
|
||||
if has_similar_in_bank:
|
||||
continue
|
||||
|
||||
if num_queries==0:
|
||||
query_images[label] = feat[None, ...]
|
||||
else:
|
||||
query_images[label] = torch.cat([query_images[label], feat[None, ...]])
|
||||
return query_images
|
||||
|
||||
def flatten_fpn_features(self, features):
|
||||
# downsample and flat fpn features for pre-select in language backbone
|
||||
return torch.cat([self.pool(f).flatten(-2,-1) for i, f in enumerate(features)], dim=2).permute(0,2,1)
|
||||
|
||||
@torch.no_grad()
|
||||
def get_labels_and_maps_from_positive_map(self, positive_map, dtype=torch.float):
|
||||
# Only for inference
|
||||
labels_in_caption=[k for k,v in positive_map.items() if len(v) !=0]
|
||||
num_labels=len(labels_in_caption)
|
||||
all_map = torch.zeros((num_labels, self.cfg.MODEL.LANGUAGE_BACKBONE.MAX_QUERY_LEN), dtype=dtype, device=self.cfg.MODEL.DEVICE)
|
||||
for j, label in enumerate(labels_in_caption):
|
||||
position=positive_map[label]
|
||||
all_map[j, position] = 1 # inplace
|
||||
all_map = all_map / (all_map.sum(-1)[:, None] + 1e-6)
|
||||
return labels_in_caption, all_map
|
||||
|
||||
def forward(self, samples: NestedTensor, targets: List = None, **kw):
|
||||
"""The forward expects a NestedTensor, which consists of:
|
||||
- samples.tensor: batched images, of shape [batch_size x 3 x H x W]
|
||||
- samples.mask: a binary mask of shape [batch_size x H x W], containing 1 on padded pixels
|
||||
|
||||
It returns a dict with the following elements:
|
||||
- "pred_logits": the classification logits (including no-object) for all queries.
|
||||
Shape= [batch_size x num_queries x num_classes]
|
||||
- "pred_boxes": The normalized boxes coordinates for all queries, represented as
|
||||
(center_x, center_y, width, height). These values are normalized in [0, 1],
|
||||
relative to the size of each individual image (disregarding possible padding).
|
||||
See PostProcess for information on how to retrieve the unnormalized bounding box.
|
||||
- "aux_outputs": Optional, only returned when auxilary losses are activated. It is a list of
|
||||
dictionnaries containing the two above keys for each decoder layer.
|
||||
"""
|
||||
if isinstance(samples, ImageList):
|
||||
image_sizes = samples.image_sizes
|
||||
samples = samples.tensors
|
||||
if targets is None:
|
||||
captions = kw["captions"]
|
||||
else:
|
||||
captions = [t.get_field("caption") for t in targets if "caption" in t.fields()]
|
||||
len(captions)
|
||||
|
||||
captions = [preprocess_caption(c) for c in captions]
|
||||
|
||||
|
||||
positive_map = kw['positive_map']
|
||||
try:
|
||||
return_backbone_features = kw['return_backbone_features']
|
||||
except:
|
||||
return_backbone_features = False
|
||||
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
if isinstance(samples, (list, torch.Tensor)):
|
||||
samples = nested_tensor_from_tensor_list(samples, image_sizes=image_sizes)
|
||||
features, poss = self.backbone(samples)
|
||||
|
||||
srcs = []
|
||||
masks = []
|
||||
for l, feat in enumerate(features):
|
||||
src, mask = feat.decompose()
|
||||
srcs.append(self.input_proj[l](src))
|
||||
masks.append(mask)
|
||||
assert mask is not None
|
||||
if self.num_feature_levels > len(srcs):
|
||||
_len_srcs = len(srcs)
|
||||
for l in range(_len_srcs, self.num_feature_levels):
|
||||
if l == _len_srcs:
|
||||
src = self.input_proj[l](features[-1].tensors)
|
||||
else:
|
||||
src = self.input_proj[l](srcs[-1])
|
||||
m = samples.mask
|
||||
mask = F.interpolate(m[None].float(), size=src.shape[-2:]).to(torch.bool)[0]
|
||||
pos_l = self.backbone[1](NestedTensor(src, mask)).to(src.dtype)
|
||||
srcs.append(src)
|
||||
masks.append(mask)
|
||||
poss.append(pos_l)
|
||||
|
||||
|
||||
# query embedding
|
||||
if self.cfg.VISION_QUERY.ENABLED:
|
||||
if self.training:
|
||||
batched_labels_in_caption=[t.get_field('labels_in_caption') for t in targets]
|
||||
batched_all_map=[t.get_field('all_map') for t in targets]
|
||||
batched_pos_category_map=[t.get_field('positive_category_map') for t in targets]
|
||||
################ BUG: batched_pos_category_map is not binary ######################
|
||||
batched_pos_labels = [t.get_field('labels') for t in targets]
|
||||
else:
|
||||
assert samples.tensors.shape[0]==1 # TODO: Only support batch size = 1 for test
|
||||
labels_in_caption, all_map = self.get_labels_and_maps_from_positive_map(positive_map, dtype=srcs[0].dtype)
|
||||
batched_labels_in_caption = [labels_in_caption]
|
||||
batched_all_map = [all_map]
|
||||
batched_pos_category_map = None
|
||||
batched_pos_labels = None
|
||||
|
||||
|
||||
query_features, query_attetion_masks, batched_has_vision_query=self.query_selector(batched_labels_in_caption, batched_all_map, batched_pos_labels)
|
||||
|
||||
vision_inputs_in_language_backbone={'vision': query_features, 'images': self.flatten_fpn_features(srcs), 'vision_attention_mask': query_attetion_masks, 'batched_pos_category_map': batched_pos_category_map}
|
||||
else:
|
||||
vision_inputs_in_language_backbone={'vision': None, 'images': None, 'vision_attention_mask': None, 'batched_pos_category_map': None}
|
||||
|
||||
|
||||
# encoder texts
|
||||
# assume each category is consist of its text tokens and one '.'
|
||||
# tokenized = self.tokenizer(captions, padding="longest", return_tensors="pt").to(
|
||||
# samples.device
|
||||
# )
|
||||
tokenized = self.tokenizer(captions, padding='max_length', return_tensors="pt").to(
|
||||
samples.device
|
||||
)
|
||||
(
|
||||
text_self_attention_masks, # each category token only attend to its own category tokens and one '.'
|
||||
position_ids, # [[0, 0, 1, 2, 0, 1, 0]]
|
||||
cate_to_token_mask_list,
|
||||
) = generate_masks_with_special_tokens_and_transfer_map(
|
||||
tokenized, self.specical_tokens, self.tokenizer
|
||||
)
|
||||
|
||||
if text_self_attention_masks.shape[1] > self.max_text_len:
|
||||
text_self_attention_masks = text_self_attention_masks[
|
||||
:, : self.max_text_len, : self.max_text_len
|
||||
]
|
||||
position_ids = position_ids[:, : self.max_text_len]
|
||||
tokenized["input_ids"] = tokenized["input_ids"][:, : self.max_text_len]
|
||||
tokenized["attention_mask"] = tokenized["attention_mask"][:, : self.max_text_len]
|
||||
tokenized["token_type_ids"] = tokenized["token_type_ids"][:, : self.max_text_len]
|
||||
|
||||
# extract text embeddings
|
||||
if self.sub_sentence_present:
|
||||
tokenized_for_encoder = {k: v for k, v in tokenized.items() if k != "attention_mask"}
|
||||
tokenized_for_encoder["attention_mask"] = text_self_attention_masks
|
||||
tokenized_for_encoder["position_ids"] = position_ids
|
||||
else:
|
||||
# import ipdb; ipdb.set_trace()
|
||||
tokenized_for_encoder = tokenized
|
||||
|
||||
tokenized_for_encoder.update(vision_inputs_in_language_backbone)
|
||||
|
||||
bert_output = self.bert(**tokenized_for_encoder) # bs, 195, 768
|
||||
|
||||
encoded_text = self.feat_map(bert_output["last_hidden_state"]) # bs, 195, d_model
|
||||
text_token_mask = tokenized.attention_mask.bool() # bs, 195
|
||||
# text_token_mask: True for nomask, False for mask
|
||||
# text_self_attention_masks: True for nomask, False for mask
|
||||
|
||||
if encoded_text.shape[1] > self.max_text_len:
|
||||
encoded_text = encoded_text[:, : self.max_text_len, :]
|
||||
text_token_mask = text_token_mask[:, : self.max_text_len]
|
||||
position_ids = position_ids[:, : self.max_text_len]
|
||||
text_self_attention_masks = text_self_attention_masks[
|
||||
:, : self.max_text_len, : self.max_text_len
|
||||
]
|
||||
|
||||
text_dict = {
|
||||
"encoded_text": encoded_text, # bs, 195, d_model
|
||||
"text_token_mask": text_token_mask, # bs, 195
|
||||
"position_ids": position_ids, # bs, 195
|
||||
"text_self_attention_masks": text_self_attention_masks, # bs, 195,195
|
||||
}
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
input_query_bbox = input_query_label = attn_mask = dn_meta = None
|
||||
hs, reference, hs_enc, ref_enc, init_box_proposal = self.transformer(
|
||||
srcs, masks, input_query_bbox, poss, input_query_label, attn_mask, text_dict
|
||||
)
|
||||
|
||||
# deformable-detr-like anchor update
|
||||
outputs_coord_list = []
|
||||
for dec_lid, (layer_ref_sig, layer_bbox_embed, layer_hs) in enumerate(
|
||||
zip(reference[:-1], self.bbox_embed, hs)
|
||||
):
|
||||
layer_delta_unsig = layer_bbox_embed(layer_hs)
|
||||
layer_outputs_unsig = layer_delta_unsig + inverse_sigmoid(layer_ref_sig)
|
||||
layer_outputs_unsig = layer_outputs_unsig.sigmoid()
|
||||
outputs_coord_list.append(layer_outputs_unsig)
|
||||
outputs_coord_list = torch.stack(outputs_coord_list)
|
||||
|
||||
# output
|
||||
outputs_class = torch.stack(
|
||||
[
|
||||
layer_cls_embed(layer_hs, text_dict)
|
||||
for layer_cls_embed, layer_hs in zip(self.class_embed, hs)
|
||||
]
|
||||
)
|
||||
if self.training:
|
||||
out = {"pred_logits": outputs_class[-1], "pred_boxes": outputs_coord_list[-1]}
|
||||
aux_outputs = [{"pred_logits": outputs_class[k], "pred_boxes": outputs_coord_list[k]} for k in range(len(outputs_class)-1)]
|
||||
out['aux_outputs'] = aux_outputs
|
||||
positive_map_ = positive_map.clone().to(outputs_class[-1].device)
|
||||
positive_map_[positive_map_>0]=1.
|
||||
|
||||
# padding to max_text_len
|
||||
text_mask = torch.full((*text_dict["text_token_mask"].shape[:-1], self.max_text_len), bool(False), device=text_dict["text_token_mask"].device)
|
||||
text_mask[..., : text_dict["text_token_mask"].shape[-1]] = text_dict["text_token_mask"]
|
||||
|
||||
losses = self.loss_evaluator(out, targets, text_mask=text_mask ,positive_map=positive_map_)
|
||||
|
||||
if self.cfg.VISION_QUERY.ENABLED:
|
||||
#### gate loss #####
|
||||
# concatenate all gates
|
||||
gates = []
|
||||
for _ ,g in bert_output['vision_query_gates'].items():
|
||||
gates = gates + g
|
||||
|
||||
num_gates=len(gates)
|
||||
loss_gate=0
|
||||
for g in gates:
|
||||
loss_gate=loss_gate+(1-torch.abs(g[0]))
|
||||
loss_gate= self.cfg.VISION_QUERY.GATE_REGULARIZATION_SCALE * loss_gate / num_gates
|
||||
if self.cfg.VISION_QUERY.GATE_REGULARIZATION:
|
||||
gate_losses = {'loss_gate': loss_gate.sum()}
|
||||
else:
|
||||
loss_gate = loss_gate.sum().detach() # Only for analysis
|
||||
gate_losses = {'loss_gate': loss_gate}
|
||||
####################
|
||||
|
||||
losses.update(gate_losses)
|
||||
return losses
|
||||
else:
|
||||
out = {"pred_logits": outputs_class[-1].sigmoid(), "pred_boxes": outputs_coord_list[-1]}
|
||||
result = self.convert_groundingdino_to_glip_output(out, positive_map, image_sizes)
|
||||
if return_backbone_features:
|
||||
return result, srcs
|
||||
return result
|
||||
|
||||
|
||||
# # for intermediate outputs
|
||||
# if self.aux_loss:
|
||||
# out['aux_outputs'] = self._set_aux_loss(outputs_class, outputs_coord_list)
|
||||
|
||||
# # for encoder output
|
||||
# if hs_enc is not None:
|
||||
# # prepare intermediate outputs
|
||||
# interm_coord = ref_enc[-1]
|
||||
# interm_class = self.transformer.enc_out_class_embed(hs_enc[-1], text_dict)
|
||||
# out['interm_outputs'] = {'pred_logits': interm_class, 'pred_boxes': interm_coord}
|
||||
# out['interm_outputs_for_matching_pre'] = {'pred_logits': interm_class, 'pred_boxes': init_box_proposal}
|
||||
|
||||
# return out
|
||||
|
||||
@torch.jit.unused
|
||||
def _set_aux_loss(self, outputs_class, outputs_coord):
|
||||
# this is a workaround to make torchscript happy, as torchscript
|
||||
# doesn't support dictionary with non-homogeneous values, such
|
||||
# as a dict having both a Tensor and a list.
|
||||
return [
|
||||
{"pred_logits": a, "pred_boxes": b}
|
||||
for a, b in zip(outputs_class[:-1], outputs_coord[:-1])
|
||||
]
|
||||
|
||||
|
||||
@MODULE_BUILD_FUNCS.registe_with_name(module_name="groundingdino")
|
||||
def build_groundingdino(args, cfg):
|
||||
|
||||
backbone = build_backbone(args)
|
||||
transformer = build_transformer(args)
|
||||
|
||||
dn_labelbook_size = args.dn_labelbook_size
|
||||
dec_pred_bbox_embed_share = args.dec_pred_bbox_embed_share
|
||||
sub_sentence_present = args.sub_sentence_present
|
||||
|
||||
model = GroundingDINO(
|
||||
backbone,
|
||||
transformer,
|
||||
num_queries=args.num_queries,
|
||||
aux_loss=True,
|
||||
iter_update=True,
|
||||
query_dim=4,
|
||||
num_feature_levels=args.num_feature_levels,
|
||||
nheads=args.nheads,
|
||||
dec_pred_bbox_embed_share=dec_pred_bbox_embed_share,
|
||||
two_stage_type=args.two_stage_type,
|
||||
two_stage_bbox_embed_share=args.two_stage_bbox_embed_share,
|
||||
two_stage_class_embed_share=args.two_stage_class_embed_share,
|
||||
num_patterns=args.num_patterns,
|
||||
dn_number=0,
|
||||
dn_box_noise_scale=args.dn_box_noise_scale,
|
||||
dn_label_noise_ratio=args.dn_label_noise_ratio,
|
||||
dn_labelbook_size=dn_labelbook_size,
|
||||
text_encoder_type=args.text_encoder_type,
|
||||
sub_sentence_present=sub_sentence_present,
|
||||
max_text_len=args.max_text_len,
|
||||
cfg=cfg,
|
||||
)
|
||||
|
||||
return model
|
||||
|
|
@ -0,0 +1,180 @@
|
|||
import copy
|
||||
import math
|
||||
from typing import List
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import nn
|
||||
from torchvision.ops.boxes import nms
|
||||
|
||||
from groundingdino_new.util import box_ops
|
||||
from groundingdino_new.util.misc import (NestedTensor, nested_tensor_from_tensor_list,
|
||||
accuracy, get_world_size, interpolate,
|
||||
is_dist_avail_and_initialized, inverse_sigmoid)
|
||||
|
||||
from groundingdino_new.models.GroundingDINO.matcher import build_matcher
|
||||
from groundingdino_new.models.GroundingDINO.utils import sigmoid_focal_loss, MLP
|
||||
from maskrcnn_benchmark.layers import SigmoidFocalLoss, IOULoss, TokenSigmoidFocalLoss
|
||||
|
||||
class SetCriterion(nn.Module):
|
||||
""" This class computes the loss for Conditional DETR.
|
||||
The process happens in two steps:
|
||||
1) we compute hungarian assignment between ground truth boxes and the outputs of the model
|
||||
2) we supervise each pair of matched ground-truth / prediction (supervise class and box)
|
||||
"""
|
||||
def __init__(self, matcher, cfg):
|
||||
""" Create the criterion.
|
||||
Parameters:
|
||||
num_classes: number of object categories, omitting the special no-object category
|
||||
matcher: module able to compute a matching between targets and proposals
|
||||
weight_dict: dict containing as key the names of the losses and as values their relative weight.
|
||||
losses: list of all the losses to be applied. See get_loss for list of available losses.
|
||||
focal_alpha: alpha in Focal Loss
|
||||
"""
|
||||
super().__init__()
|
||||
# self.num_classes = num_classes
|
||||
self.matcher = matcher
|
||||
self.weight_dict = {'loss_ce': cfg.GROUNDINGDINO.loss_ce_coef,'loss_bbox': cfg.GROUNDINGDINO.loss_bbox_coef,'loss_giou': cfg.GROUNDINGDINO.loss_giou_coef}
|
||||
self.losses = ['labels', 'boxes']
|
||||
self.token_loss_func = TokenSigmoidFocalLoss(cfg.MODEL.DYHEAD.FUSE_CONFIG.TOKEN_ALPHA,
|
||||
cfg.MODEL.DYHEAD.FUSE_CONFIG.TOKEN_GAMMA)
|
||||
# self.focal_alpha = focal_alpha
|
||||
|
||||
def loss_labels(self, outputs, targets, indices, num_boxes, text_mask, positive_map):
|
||||
"""Classification loss (Binary focal loss)
|
||||
targets dicts must contain the key "labels" containing a tensor of dim [nb_target_boxes]
|
||||
"""
|
||||
assert 'pred_logits' in outputs
|
||||
src_logits = outputs['pred_logits']
|
||||
|
||||
positive_map_per_image = positive_map.split([len(t) for t in targets])
|
||||
|
||||
idx = self._get_src_permutation_idx(indices)
|
||||
# target_classes_o = torch.cat([t["labels"][J] for t, (_, J) in zip(targets, indices)])
|
||||
|
||||
positive_map_per_image_o = torch.cat([pos_map[J] for pos_map, (_, J) in zip(positive_map_per_image, indices)])
|
||||
target_classes = torch.zeros(src_logits.shape, dtype=src_logits.dtype, layout=src_logits.layout, device=src_logits.device)
|
||||
target_classes[idx]=positive_map_per_image_o
|
||||
|
||||
unmatched_labels = torch.zeros(target_classes.shape[-1], device=target_classes.device)
|
||||
unmatched_labels[-1] = 1.
|
||||
target_classes[target_classes.sum(-1)==0] = unmatched_labels
|
||||
|
||||
dot_product_token_loss = self.token_loss_func(src_logits,
|
||||
target_classes, text_masks=text_mask,
|
||||
version="binary") / num_boxes
|
||||
|
||||
|
||||
losses = {'loss_ce': dot_product_token_loss}
|
||||
return losses
|
||||
|
||||
def loss_boxes(self, outputs, targets, indices, num_boxes):
|
||||
"""Compute the losses related to the bounding boxes, the L1 regression loss and the GIoU loss
|
||||
targets dicts must contain the key "boxes" containing a tensor of dim [nb_target_boxes, 4]
|
||||
The target boxes are expected in format (center_x, center_y, w, h), normalized by the image size.
|
||||
"""
|
||||
assert 'pred_boxes' in outputs
|
||||
idx = self._get_src_permutation_idx(indices)
|
||||
src_boxes = outputs['pred_boxes'][idx]
|
||||
target_boxes = torch.cat([t.get_field('normed_cxcy_boxes')[i] for t, (_, i) in zip(targets, indices)], dim=0)
|
||||
|
||||
loss_bbox = F.l1_loss(src_boxes, target_boxes, reduction='none')
|
||||
|
||||
losses = {}
|
||||
losses['loss_bbox'] = loss_bbox.sum() / num_boxes
|
||||
|
||||
loss_giou = 1 - torch.diag(box_ops.generalized_box_iou(
|
||||
box_ops.box_cxcywh_to_xyxy(src_boxes),
|
||||
box_ops.box_cxcywh_to_xyxy(target_boxes)))
|
||||
losses['loss_giou'] = loss_giou.sum() / num_boxes
|
||||
|
||||
|
||||
return losses
|
||||
|
||||
|
||||
def _get_src_permutation_idx(self, indices):
|
||||
# permute predictions following indices
|
||||
batch_idx = torch.cat([torch.full_like(src, i) for i, (src, _) in enumerate(indices)])
|
||||
src_idx = torch.cat([src for (src, _) in indices])
|
||||
return batch_idx, src_idx
|
||||
|
||||
def _get_tgt_permutation_idx(self, indices):
|
||||
# permute targets following indices
|
||||
batch_idx = torch.cat([torch.full_like(tgt, i) for i, (_, tgt) in enumerate(indices)])
|
||||
tgt_idx = torch.cat([tgt for (_, tgt) in indices])
|
||||
return batch_idx, tgt_idx
|
||||
|
||||
def get_loss(self, loss, outputs, targets, indices, num_boxes, **kwargs):
|
||||
loss_map = {
|
||||
'labels': self.loss_labels,
|
||||
'boxes': self.loss_boxes,
|
||||
}
|
||||
assert loss in loss_map, f'do you really want to compute {loss} loss?'
|
||||
return loss_map[loss](outputs, targets, indices, num_boxes, **kwargs)
|
||||
|
||||
def forward(self, outputs, targets, return_indices=False, text_mask=None, positive_map=None):
|
||||
""" This performs the loss computation.
|
||||
Parameters:
|
||||
outputs: dict of tensors, see the output specification of the model for the format
|
||||
targets: list of dicts, such that len(targets) == batch_size.
|
||||
The expected keys in each dict depends on the losses applied, see each loss' doc
|
||||
|
||||
return_indices: used for vis. if True, the layer0-5 indices will be returned as well.
|
||||
"""
|
||||
outputs_without_aux = {k: v for k, v in outputs.items() if k != 'aux_outputs'}
|
||||
device=next(iter(outputs.values())).device
|
||||
indices = self.matcher(outputs_without_aux, targets, positive_map)
|
||||
|
||||
if return_indices:
|
||||
indices0_copy = indices
|
||||
indices_list = []
|
||||
|
||||
# Compute the average number of target boxes accross all nodes, for normalization purposes
|
||||
num_boxes = len(positive_map)
|
||||
num_boxes = torch.as_tensor([num_boxes], dtype=torch.float, device=device)
|
||||
if is_dist_avail_and_initialized():
|
||||
torch.distributed.all_reduce(num_boxes)
|
||||
num_boxes = torch.clamp(num_boxes / get_world_size(), min=1).item()
|
||||
|
||||
# Compute all the requested losses
|
||||
losses = {}
|
||||
|
||||
for loss in self.losses:
|
||||
if 'labels' in loss:
|
||||
losses.update(self.get_loss(loss, outputs, targets, indices, num_boxes, text_mask=text_mask, positive_map=positive_map))
|
||||
else:
|
||||
losses.update(self.get_loss(loss, outputs, targets, indices, num_boxes))
|
||||
|
||||
# In case of auxiliary losses, we repeat this process with the output of each intermediate layer.
|
||||
if 'aux_outputs' in outputs:
|
||||
for idx, aux_outputs in enumerate(outputs['aux_outputs']):
|
||||
indices = self.matcher(aux_outputs, targets, positive_map)
|
||||
if return_indices:
|
||||
indices_list.append(indices)
|
||||
for loss in self.losses:
|
||||
if loss == 'masks':
|
||||
# Intermediate masks losses are too costly to compute, we ignore them.
|
||||
continue
|
||||
kwargs = {}
|
||||
|
||||
if 'labels' in loss:
|
||||
l_dict = self.get_loss(loss, aux_outputs, targets, indices, num_boxes, text_mask=text_mask, positive_map=positive_map, **kwargs)
|
||||
else:
|
||||
l_dict = self.get_loss(loss, aux_outputs, targets, indices, num_boxes, **kwargs)
|
||||
l_dict = {k + f'_{idx}': v for k, v in l_dict.items()}
|
||||
losses.update(l_dict)
|
||||
|
||||
new_losses = {}
|
||||
for k,v in losses.items():
|
||||
for name, weight in self.weight_dict.items():
|
||||
if name in k:
|
||||
new_losses[k] = v * weight
|
||||
losses.update(new_losses)
|
||||
|
||||
|
||||
if return_indices:
|
||||
indices_list.append(indices0_copy)
|
||||
return losses, indices_list
|
||||
|
||||
return losses
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,182 @@
|
|||
import torch, os
|
||||
from torch import nn
|
||||
from scipy.optimize import linear_sum_assignment
|
||||
|
||||
from groundingdino_new.util.box_ops import box_cxcywh_to_xyxy, generalized_box_iou
|
||||
|
||||
|
||||
class HungarianMatcher(nn.Module):
|
||||
"""This class computes an assignment between the targets and the predictions of the network
|
||||
For efficiency reasons, the targets don't include the no_object. Because of this, in general,
|
||||
there are more predictions than targets. In this case, we do a 1-to-1 matching of the best predictions,
|
||||
while the others are un-matched (and thus treated as non-objects).
|
||||
"""
|
||||
|
||||
def __init__(self, cost_class: float = 1, cost_bbox: float = 1, cost_giou: float = 1, focal_alpha = 0.25):
|
||||
"""Creates the matcher
|
||||
Params:
|
||||
cost_class: This is the relative weight of the classification error in the matching cost
|
||||
cost_bbox: This is the relative weight of the L1 error of the bounding box coordinates in the matching cost
|
||||
cost_giou: This is the relative weight of the giou loss of the bounding box in the matching cost
|
||||
"""
|
||||
super().__init__()
|
||||
self.cost_class = cost_class
|
||||
self.cost_bbox = cost_bbox
|
||||
self.cost_giou = cost_giou
|
||||
assert cost_class != 0 or cost_bbox != 0 or cost_giou != 0, "all costs cant be 0"
|
||||
|
||||
self.focal_alpha = focal_alpha
|
||||
|
||||
@torch.no_grad()
|
||||
def forward(self, outputs, targets, positive_map):
|
||||
""" Performs the matching
|
||||
Params:
|
||||
outputs: This is a dict that contains at least these entries:
|
||||
"pred_logits": Tensor of dim [batch_size, num_queries, num_classes] with the classification logits
|
||||
"pred_boxes": Tensor of dim [batch_size, num_queries, 4] with the predicted box coordinates
|
||||
targets: This is a list of targets (len(targets) = batch_size), where each target is a dict containing:
|
||||
"labels": Tensor of dim [num_target_boxes] (where num_target_boxes is the number of ground-truth
|
||||
objects in the target) containing the class labels
|
||||
"boxes": Tensor of dim [num_target_boxes, 4] containing the target box coordinates
|
||||
Returns:
|
||||
A list of size batch_size, containing tuples of (index_i, index_j) where:
|
||||
- index_i is the indices of the selected predictions (in order)
|
||||
- index_j is the indices of the corresponding selected targets (in order)
|
||||
For each batch element, it holds:
|
||||
len(index_i) = len(index_j) = min(num_queries, num_target_boxes)
|
||||
"""
|
||||
|
||||
bs, num_queries = outputs["pred_logits"].shape[:2]
|
||||
|
||||
# We flatten to compute the cost matrices in a batch
|
||||
out_prob = outputs["pred_logits"].flatten(0, 1).sigmoid() # [batch_size * num_queries, num_classes]
|
||||
out_bbox = outputs["pred_boxes"].flatten(0, 1) # [batch_size * num_queries, 4]
|
||||
|
||||
# Also concat the target labels and boxes
|
||||
# tgt_ids = torch.cat([v["labels"] for v in targets])
|
||||
tgt_ids = (positive_map>0)
|
||||
tgt_bbox = torch.cat([v.get_field('normed_cxcy_boxes') for v in targets])
|
||||
|
||||
# Compute the classification cost.
|
||||
alpha = self.focal_alpha
|
||||
gamma = 2.0
|
||||
neg_cost_class = (1 - alpha) * (out_prob ** gamma) * (-(1 - out_prob + 1e-8).log())
|
||||
pos_cost_class = alpha * ((1 - out_prob) ** gamma) * (-(out_prob + 1e-8).log())
|
||||
|
||||
# cost_class = pos_cost_class[:, tgt_ids] - neg_cost_class[:, tgt_ids]
|
||||
cost_class = []
|
||||
for pos_m in tgt_ids:
|
||||
cost_class.append((pos_cost_class[:, pos_m] - neg_cost_class[:, pos_m]).mean(-1))
|
||||
cost_class = torch.stack(cost_class).transpose(1,0)
|
||||
|
||||
|
||||
|
||||
# Compute the L1 cost between boxes
|
||||
cost_bbox = torch.cdist(out_bbox, tgt_bbox, p=1)
|
||||
|
||||
# Compute the giou cost betwen boxes
|
||||
cost_giou = -generalized_box_iou(box_cxcywh_to_xyxy(out_bbox), box_cxcywh_to_xyxy(tgt_bbox))
|
||||
|
||||
# Final cost matrix
|
||||
C = self.cost_bbox * cost_bbox + self.cost_class * cost_class + self.cost_giou * cost_giou
|
||||
C = C.view(bs, num_queries, -1).cpu()
|
||||
|
||||
sizes = [len(v.get_field('normed_cxcy_boxes')) for v in targets]
|
||||
indices = [linear_sum_assignment(c[i]) for i, c in enumerate(C.split(sizes, -1))]
|
||||
return [(torch.as_tensor(i, dtype=torch.int64), torch.as_tensor(j, dtype=torch.int64)) for i, j in indices]
|
||||
|
||||
|
||||
class SimpleMinsumMatcher(nn.Module):
|
||||
"""This class computes an assignment between the targets and the predictions of the network
|
||||
For efficiency reasons, the targets don't include the no_object. Because of this, in general,
|
||||
there are more predictions than targets. In this case, we do a 1-to-1 matching of the best predictions,
|
||||
while the others are un-matched (and thus treated as non-objects).
|
||||
"""
|
||||
|
||||
def __init__(self, cost_class: float = 1, cost_bbox: float = 1, cost_giou: float = 1, focal_alpha = 0.25):
|
||||
"""Creates the matcher
|
||||
Params:
|
||||
cost_class: This is the relative weight of the classification error in the matching cost
|
||||
cost_bbox: This is the relative weight of the L1 error of the bounding box coordinates in the matching cost
|
||||
cost_giou: This is the relative weight of the giou loss of the bounding box in the matching cost
|
||||
"""
|
||||
super().__init__()
|
||||
self.cost_class = cost_class
|
||||
self.cost_bbox = cost_bbox
|
||||
self.cost_giou = cost_giou
|
||||
assert cost_class != 0 or cost_bbox != 0 or cost_giou != 0, "all costs cant be 0"
|
||||
|
||||
self.focal_alpha = focal_alpha
|
||||
|
||||
@torch.no_grad()
|
||||
def forward(self, outputs, targets):
|
||||
""" Performs the matching
|
||||
Params:
|
||||
outputs: This is a dict that contains at least these entries:
|
||||
"pred_logits": Tensor of dim [batch_size, num_queries, num_classes] with the classification logits
|
||||
"pred_boxes": Tensor of dim [batch_size, num_queries, 4] with the predicted box coordinates
|
||||
targets: This is a list of targets (len(targets) = batch_size), where each target is a dict containing:
|
||||
"labels": Tensor of dim [num_target_boxes] (where num_target_boxes is the number of ground-truth
|
||||
objects in the target) containing the class labels
|
||||
"boxes": Tensor of dim [num_target_boxes, 4] containing the target box coordinates
|
||||
Returns:
|
||||
A list of size batch_size, containing tuples of (index_i, index_j) where:
|
||||
- index_i is the indices of the selected predictions (in order)
|
||||
- index_j is the indices of the corresponding selected targets (in order)
|
||||
For each batch element, it holds:
|
||||
len(index_i) = len(index_j) = min(num_queries, num_target_boxes)
|
||||
"""
|
||||
|
||||
bs, num_queries = outputs["pred_logits"].shape[:2]
|
||||
|
||||
# We flatten to compute the cost matrices in a batch
|
||||
out_prob = outputs["pred_logits"].flatten(0, 1).sigmoid() # [batch_size * num_queries, num_classes]
|
||||
out_bbox = outputs["pred_boxes"].flatten(0, 1) # [batch_size * num_queries, 4]
|
||||
|
||||
# Also concat the target labels and boxes
|
||||
tgt_ids = torch.cat([v["labels"] for v in targets])
|
||||
tgt_bbox = torch.cat([v["boxes"] for v in targets])
|
||||
|
||||
# Compute the classification cost.
|
||||
alpha = self.focal_alpha
|
||||
gamma = 2.0
|
||||
neg_cost_class = (1 - alpha) * (out_prob ** gamma) * (-(1 - out_prob + 1e-8).log())
|
||||
pos_cost_class = alpha * ((1 - out_prob) ** gamma) * (-(out_prob + 1e-8).log())
|
||||
cost_class = pos_cost_class[:, tgt_ids] - neg_cost_class[:, tgt_ids]
|
||||
|
||||
# Compute the L1 cost between boxes
|
||||
cost_bbox = torch.cdist(out_bbox, tgt_bbox, p=1)
|
||||
|
||||
# Compute the giou cost betwen boxes
|
||||
cost_giou = -generalized_box_iou(box_cxcywh_to_xyxy(out_bbox), box_cxcywh_to_xyxy(tgt_bbox))
|
||||
|
||||
# Final cost matrix
|
||||
C = self.cost_bbox * cost_bbox + self.cost_class * cost_class + self.cost_giou * cost_giou
|
||||
C = C.view(bs, num_queries, -1)
|
||||
|
||||
sizes = [len(v["boxes"]) for v in targets]
|
||||
indices = []
|
||||
device = C.device
|
||||
for i, (c, _size) in enumerate(zip(C.split(sizes, -1), sizes)):
|
||||
weight_mat = c[i]
|
||||
idx_i = weight_mat.min(0)[1]
|
||||
idx_j = torch.arange(_size).to(device)
|
||||
indices.append((idx_i, idx_j))
|
||||
|
||||
return [(torch.as_tensor(i, dtype=torch.int64), torch.as_tensor(j, dtype=torch.int64)) for i, j in indices]
|
||||
|
||||
|
||||
def build_matcher(args):
|
||||
assert args.matcher_type in ['HungarianMatcher', 'SimpleMinsumMatcher'], "Unknown args.matcher_type: {}".format(args.matcher_type)
|
||||
if args.matcher_type == 'HungarianMatcher':
|
||||
return HungarianMatcher(
|
||||
cost_class=args.set_cost_class, cost_bbox=args.set_cost_bbox, cost_giou=args.set_cost_giou,
|
||||
focal_alpha=args.focal_alpha
|
||||
)
|
||||
elif args.matcher_type == 'SimpleMinsumMatcher':
|
||||
return SimpleMinsumMatcher(
|
||||
cost_class=args.set_cost_class, cost_bbox=args.set_cost_bbox, cost_giou=args.set_cost_giou,
|
||||
focal_alpha=args.focal_alpha
|
||||
)
|
||||
else:
|
||||
raise NotImplementedError("Unknown args.matcher_type: {}".format(args.matcher_type))
|
||||
|
|
@ -0,0 +1,413 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Deformable DETR
|
||||
# Copyright (c) 2020 SenseTime. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------------------------------
|
||||
# Modified from:
|
||||
# https://github.com/fundamentalvision/Deformable-DETR/blob/main/models/ops/functions/ms_deform_attn_func.py
|
||||
# https://github.com/fundamentalvision/Deformable-DETR/blob/main/models/ops/modules/ms_deform_attn.py
|
||||
# https://github.com/open-mmlab/mmcv/blob/master/mmcv/ops/multi_scale_deform_attn.py
|
||||
# ------------------------------------------------------------------------------------------------
|
||||
|
||||
import math
|
||||
import warnings
|
||||
from typing import Optional
|
||||
|
||||
import torch
|
||||
import torch.nn as nn
|
||||
import torch.nn.functional as F
|
||||
from torch.autograd import Function
|
||||
from torch.autograd.function import once_differentiable
|
||||
from torch.nn.init import constant_, xavier_uniform_
|
||||
|
||||
try:
|
||||
from groundingdino_new import _C
|
||||
except:
|
||||
warnings.warn("Failed to load custom C++ ops. Running on CPU mode Only!")
|
||||
|
||||
|
||||
# helpers
|
||||
def _is_power_of_2(n):
|
||||
if (not isinstance(n, int)) or (n < 0):
|
||||
raise ValueError("invalid input for _is_power_of_2: {} (type: {})".format(n, type(n)))
|
||||
return (n & (n - 1) == 0) and n != 0
|
||||
|
||||
|
||||
class MultiScaleDeformableAttnFunction(Function):
|
||||
@staticmethod
|
||||
def forward(
|
||||
ctx,
|
||||
value,
|
||||
value_spatial_shapes,
|
||||
value_level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
im2col_step,
|
||||
):
|
||||
ctx.im2col_step = im2col_step
|
||||
output = _C.ms_deform_attn_forward(
|
||||
value,
|
||||
value_spatial_shapes,
|
||||
value_level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
ctx.im2col_step,
|
||||
)
|
||||
ctx.save_for_backward(
|
||||
value,
|
||||
value_spatial_shapes,
|
||||
value_level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
)
|
||||
return output
|
||||
|
||||
@staticmethod
|
||||
@once_differentiable
|
||||
def backward(ctx, grad_output):
|
||||
(
|
||||
value,
|
||||
value_spatial_shapes,
|
||||
value_level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
) = ctx.saved_tensors
|
||||
grad_value, grad_sampling_loc, grad_attn_weight = _C.ms_deform_attn_backward(
|
||||
value,
|
||||
value_spatial_shapes,
|
||||
value_level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
grad_output,
|
||||
ctx.im2col_step,
|
||||
)
|
||||
|
||||
return grad_value, None, None, grad_sampling_loc, grad_attn_weight, None
|
||||
|
||||
|
||||
def multi_scale_deformable_attn_pytorch(
|
||||
value: torch.Tensor,
|
||||
value_spatial_shapes: torch.Tensor,
|
||||
sampling_locations: torch.Tensor,
|
||||
attention_weights: torch.Tensor,
|
||||
) -> torch.Tensor:
|
||||
|
||||
bs, _, num_heads, embed_dims = value.shape
|
||||
_, num_queries, num_heads, num_levels, num_points, _ = sampling_locations.shape
|
||||
value_list = value.split([H_ * W_ for H_, W_ in value_spatial_shapes], dim=1)
|
||||
sampling_grids = 2 * sampling_locations - 1
|
||||
sampling_value_list = []
|
||||
for level, (H_, W_) in enumerate(value_spatial_shapes):
|
||||
# bs, H_*W_, num_heads, embed_dims ->
|
||||
# bs, H_*W_, num_heads*embed_dims ->
|
||||
# bs, num_heads*embed_dims, H_*W_ ->
|
||||
# bs*num_heads, embed_dims, H_, W_
|
||||
value_l_ = (
|
||||
value_list[level].flatten(2).transpose(1, 2).reshape(bs * num_heads, embed_dims, H_, W_)
|
||||
)
|
||||
# bs, num_queries, num_heads, num_points, 2 ->
|
||||
# bs, num_heads, num_queries, num_points, 2 ->
|
||||
# bs*num_heads, num_queries, num_points, 2
|
||||
sampling_grid_l_ = sampling_grids[:, :, :, level].transpose(1, 2).flatten(0, 1)
|
||||
# bs*num_heads, embed_dims, num_queries, num_points
|
||||
sampling_value_l_ = F.grid_sample(
|
||||
value_l_, sampling_grid_l_, mode="bilinear", padding_mode="zeros", align_corners=False
|
||||
)
|
||||
sampling_value_list.append(sampling_value_l_)
|
||||
# (bs, num_queries, num_heads, num_levels, num_points) ->
|
||||
# (bs, num_heads, num_queries, num_levels, num_points) ->
|
||||
# (bs, num_heads, 1, num_queries, num_levels*num_points)
|
||||
attention_weights = attention_weights.transpose(1, 2).reshape(
|
||||
bs * num_heads, 1, num_queries, num_levels * num_points
|
||||
)
|
||||
output = (
|
||||
(torch.stack(sampling_value_list, dim=-2).flatten(-2) * attention_weights)
|
||||
.sum(-1)
|
||||
.view(bs, num_heads * embed_dims, num_queries)
|
||||
)
|
||||
return output.transpose(1, 2).contiguous()
|
||||
|
||||
|
||||
class MultiScaleDeformableAttention(nn.Module):
|
||||
"""Multi-Scale Deformable Attention Module used in Deformable-DETR
|
||||
|
||||
`Deformable DETR: Deformable Transformers for End-to-End Object Detection.
|
||||
<https://arxiv.org/pdf/2010.04159.pdf>`_.
|
||||
|
||||
Args:
|
||||
embed_dim (int): The embedding dimension of Attention. Default: 256.
|
||||
num_heads (int): The number of attention heads. Default: 8.
|
||||
num_levels (int): The number of feature map used in Attention. Default: 4.
|
||||
num_points (int): The number of sampling points for each query
|
||||
in each head. Default: 4.
|
||||
img2col_steps (int): The step used in image_to_column. Defualt: 64.
|
||||
dropout (float): Dropout layer used in output. Default: 0.1.
|
||||
batch_first (bool): if ``True``, then the input and output tensor will be
|
||||
provided as `(bs, n, embed_dim)`. Default: False. `(n, bs, embed_dim)`
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
embed_dim: int = 256,
|
||||
num_heads: int = 8,
|
||||
num_levels: int = 4,
|
||||
num_points: int = 4,
|
||||
img2col_step: int = 64,
|
||||
batch_first: bool = False,
|
||||
):
|
||||
super().__init__()
|
||||
if embed_dim % num_heads != 0:
|
||||
raise ValueError(
|
||||
"embed_dim must be divisible by num_heads, but got {} and {}".format(
|
||||
embed_dim, num_heads
|
||||
)
|
||||
)
|
||||
head_dim = embed_dim // num_heads
|
||||
|
||||
self.batch_first = batch_first
|
||||
|
||||
if not _is_power_of_2(head_dim):
|
||||
warnings.warn(
|
||||
"""
|
||||
You'd better set d_model in MSDeformAttn to make sure that
|
||||
each dim of the attention head a power of 2, which is more efficient.
|
||||
"""
|
||||
)
|
||||
|
||||
self.im2col_step = img2col_step
|
||||
self.embed_dim = embed_dim
|
||||
self.num_heads = num_heads
|
||||
self.num_levels = num_levels
|
||||
self.num_points = num_points
|
||||
self.sampling_offsets = nn.Linear(embed_dim, num_heads * num_levels * num_points * 2)
|
||||
self.attention_weights = nn.Linear(embed_dim, num_heads * num_levels * num_points)
|
||||
self.value_proj = nn.Linear(embed_dim, embed_dim)
|
||||
self.output_proj = nn.Linear(embed_dim, embed_dim)
|
||||
|
||||
self.init_weights()
|
||||
|
||||
def _reset_parameters(self):
|
||||
return self.init_weights()
|
||||
|
||||
def init_weights(self):
|
||||
"""
|
||||
Default initialization for Parameters of Module.
|
||||
"""
|
||||
constant_(self.sampling_offsets.weight.data, 0.0)
|
||||
thetas = torch.arange(self.num_heads, dtype=torch.float32) * (
|
||||
2.0 * math.pi / self.num_heads
|
||||
)
|
||||
grid_init = torch.stack([thetas.cos(), thetas.sin()], -1)
|
||||
grid_init = (
|
||||
(grid_init / grid_init.abs().max(-1, keepdim=True)[0])
|
||||
.view(self.num_heads, 1, 1, 2)
|
||||
.repeat(1, self.num_levels, self.num_points, 1)
|
||||
)
|
||||
for i in range(self.num_points):
|
||||
grid_init[:, :, i, :] *= i + 1
|
||||
with torch.no_grad():
|
||||
self.sampling_offsets.bias = nn.Parameter(grid_init.view(-1))
|
||||
constant_(self.attention_weights.weight.data, 0.0)
|
||||
constant_(self.attention_weights.bias.data, 0.0)
|
||||
xavier_uniform_(self.value_proj.weight.data)
|
||||
constant_(self.value_proj.bias.data, 0.0)
|
||||
xavier_uniform_(self.output_proj.weight.data)
|
||||
constant_(self.output_proj.bias.data, 0.0)
|
||||
|
||||
def freeze_sampling_offsets(self):
|
||||
print("Freeze sampling offsets")
|
||||
self.sampling_offsets.weight.requires_grad = False
|
||||
self.sampling_offsets.bias.requires_grad = False
|
||||
|
||||
def freeze_attention_weights(self):
|
||||
print("Freeze attention weights")
|
||||
self.attention_weights.weight.requires_grad = False
|
||||
self.attention_weights.bias.requires_grad = False
|
||||
|
||||
def forward(
|
||||
self,
|
||||
query: torch.Tensor,
|
||||
key: Optional[torch.Tensor] = None,
|
||||
value: Optional[torch.Tensor] = None,
|
||||
query_pos: Optional[torch.Tensor] = None,
|
||||
key_padding_mask: Optional[torch.Tensor] = None,
|
||||
reference_points: Optional[torch.Tensor] = None,
|
||||
spatial_shapes: Optional[torch.Tensor] = None,
|
||||
level_start_index: Optional[torch.Tensor] = None,
|
||||
**kwargs
|
||||
) -> torch.Tensor:
|
||||
|
||||
"""Forward Function of MultiScaleDeformableAttention
|
||||
|
||||
Args:
|
||||
query (torch.Tensor): Query embeddings with shape
|
||||
`(num_query, bs, embed_dim)`
|
||||
key (torch.Tensor): Key embeddings with shape
|
||||
`(num_key, bs, embed_dim)`
|
||||
value (torch.Tensor): Value embeddings with shape
|
||||
`(num_key, bs, embed_dim)`
|
||||
query_pos (torch.Tensor): The position embedding for `query`. Default: None.
|
||||
key_padding_mask (torch.Tensor): ByteTensor for `query`, with shape `(bs, num_key)`,
|
||||
indicating which elements within `key` to be ignored in attention.
|
||||
reference_points (torch.Tensor): The normalized reference points
|
||||
with shape `(bs, num_query, num_levels, 2)`,
|
||||
all elements is range in [0, 1], top-left (0, 0),
|
||||
bottom-right (1, 1), including padding are.
|
||||
or `(N, Length_{query}, num_levels, 4)`, add additional
|
||||
two dimensions `(h, w)` to form reference boxes.
|
||||
spatial_shapes (torch.Tensor): Spatial shape of features in different levels.
|
||||
With shape `(num_levels, 2)`, last dimension represents `(h, w)`.
|
||||
level_start_index (torch.Tensor): The start index of each level. A tensor with
|
||||
shape `(num_levels, )` which can be represented as
|
||||
`[0, h_0 * w_0, h_0 * w_0 + h_1 * w_1, ...]`.
|
||||
|
||||
Returns:
|
||||
torch.Tensor: forward results with shape `(num_query, bs, embed_dim)`
|
||||
"""
|
||||
|
||||
if value is None:
|
||||
value = query
|
||||
|
||||
if query_pos is not None:
|
||||
query = query + query_pos
|
||||
|
||||
if not self.batch_first:
|
||||
# change to (bs, num_query ,embed_dims)
|
||||
query = query.permute(1, 0, 2)
|
||||
value = value.permute(1, 0, 2)
|
||||
|
||||
bs, num_query, _ = query.shape
|
||||
bs, num_value, _ = value.shape
|
||||
|
||||
assert (spatial_shapes[:, 0] * spatial_shapes[:, 1]).sum() == num_value
|
||||
|
||||
value = self.value_proj(value)
|
||||
if key_padding_mask is not None:
|
||||
value = value.masked_fill(key_padding_mask[..., None], float(0))
|
||||
value = value.view(bs, num_value, self.num_heads, -1)
|
||||
sampling_offsets = self.sampling_offsets(query).view(
|
||||
bs, num_query, self.num_heads, self.num_levels, self.num_points, 2
|
||||
)
|
||||
attention_weights = self.attention_weights(query).view(
|
||||
bs, num_query, self.num_heads, self.num_levels * self.num_points
|
||||
)
|
||||
attention_weights = attention_weights.softmax(-1)
|
||||
attention_weights = attention_weights.view(
|
||||
bs,
|
||||
num_query,
|
||||
self.num_heads,
|
||||
self.num_levels,
|
||||
self.num_points,
|
||||
)
|
||||
|
||||
# bs, num_query, num_heads, num_levels, num_points, 2
|
||||
if reference_points.shape[-1] == 2:
|
||||
offset_normalizer = torch.stack([spatial_shapes[..., 1], spatial_shapes[..., 0]], -1)
|
||||
sampling_locations = (
|
||||
reference_points[:, :, None, :, None, :]
|
||||
+ sampling_offsets / offset_normalizer[None, None, None, :, None, :]
|
||||
)
|
||||
elif reference_points.shape[-1] == 4:
|
||||
sampling_locations = (
|
||||
reference_points[:, :, None, :, None, :2]
|
||||
+ sampling_offsets
|
||||
/ self.num_points
|
||||
* reference_points[:, :, None, :, None, 2:]
|
||||
* 0.5
|
||||
)
|
||||
else:
|
||||
raise ValueError(
|
||||
"Last dim of reference_points must be 2 or 4, but get {} instead.".format(
|
||||
reference_points.shape[-1]
|
||||
)
|
||||
)
|
||||
|
||||
if torch.cuda.is_available() and value.is_cuda:
|
||||
halffloat = False
|
||||
if value.dtype == torch.float16:
|
||||
halffloat = True
|
||||
value = value.float()
|
||||
sampling_locations = sampling_locations.float()
|
||||
attention_weights = attention_weights.float()
|
||||
|
||||
output = MultiScaleDeformableAttnFunction.apply(
|
||||
value,
|
||||
spatial_shapes,
|
||||
level_start_index,
|
||||
sampling_locations,
|
||||
attention_weights,
|
||||
self.im2col_step,
|
||||
)
|
||||
|
||||
if halffloat:
|
||||
output = output.half()
|
||||
else:
|
||||
output = multi_scale_deformable_attn_pytorch(
|
||||
value, spatial_shapes, sampling_locations, attention_weights
|
||||
)
|
||||
|
||||
output = self.output_proj(output)
|
||||
|
||||
if not self.batch_first:
|
||||
output = output.permute(1, 0, 2)
|
||||
|
||||
return output
|
||||
|
||||
|
||||
def create_dummy_class(klass, dependency, message=""):
|
||||
"""
|
||||
When a dependency of a class is not available, create a dummy class which throws ImportError
|
||||
when used.
|
||||
|
||||
Args:
|
||||
klass (str): name of the class.
|
||||
dependency (str): name of the dependency.
|
||||
message: extra message to print
|
||||
Returns:
|
||||
class: a class object
|
||||
"""
|
||||
err = "Cannot import '{}', therefore '{}' is not available.".format(dependency, klass)
|
||||
if message:
|
||||
err = err + " " + message
|
||||
|
||||
class _DummyMetaClass(type):
|
||||
# throw error on class attribute access
|
||||
def __getattr__(_, __): # noqa: B902
|
||||
raise ImportError(err)
|
||||
|
||||
class _Dummy(object, metaclass=_DummyMetaClass):
|
||||
# throw error on constructor
|
||||
def __init__(self, *args, **kwargs):
|
||||
raise ImportError(err)
|
||||
|
||||
return _Dummy
|
||||
|
||||
|
||||
def create_dummy_func(func, dependency, message=""):
|
||||
"""
|
||||
When a dependency of a function is not available, create a dummy function which throws
|
||||
ImportError when used.
|
||||
|
||||
Args:
|
||||
func (str): name of the function.
|
||||
dependency (str or list[str]): name(s) of the dependency.
|
||||
message: extra message to print
|
||||
Returns:
|
||||
function: a function object
|
||||
"""
|
||||
err = "Cannot import '{}', therefore '{}' is not available.".format(dependency, func)
|
||||
if message:
|
||||
err = err + " " + message
|
||||
|
||||
if isinstance(dependency, (list, tuple)):
|
||||
dependency = ",".join(dependency)
|
||||
|
||||
def _dummy(*args, **kwargs):
|
||||
raise ImportError(err)
|
||||
|
||||
return _dummy
|
||||
|
|
@ -0,0 +1,959 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# DINO
|
||||
# Copyright (c) 2022 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Conditional DETR Transformer class.
|
||||
# Copyright (c) 2021 Microsoft. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Modified from DETR (https://github.com/facebookresearch/detr)
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved.
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
from typing import Optional
|
||||
|
||||
import torch
|
||||
import torch.utils.checkpoint as checkpoint
|
||||
from torch import Tensor, nn
|
||||
|
||||
from groundingdino_new.util.misc import inverse_sigmoid
|
||||
|
||||
from .fuse_modules import BiAttentionBlock
|
||||
from .ms_deform_attn import MultiScaleDeformableAttention as MSDeformAttn
|
||||
from .transformer_vanilla import TransformerEncoderLayer
|
||||
from .utils import (
|
||||
MLP,
|
||||
_get_activation_fn,
|
||||
_get_clones,
|
||||
gen_encoder_output_proposals,
|
||||
gen_sineembed_for_position,
|
||||
get_sine_pos_embed,
|
||||
)
|
||||
|
||||
|
||||
class Transformer(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
d_model=256,
|
||||
nhead=8,
|
||||
num_queries=300,
|
||||
num_encoder_layers=6,
|
||||
num_unicoder_layers=0,
|
||||
num_decoder_layers=6,
|
||||
dim_feedforward=2048,
|
||||
dropout=0.0,
|
||||
activation="relu",
|
||||
normalize_before=False,
|
||||
return_intermediate_dec=False,
|
||||
query_dim=4,
|
||||
num_patterns=0,
|
||||
# for deformable encoder
|
||||
num_feature_levels=1,
|
||||
enc_n_points=4,
|
||||
dec_n_points=4,
|
||||
# init query
|
||||
learnable_tgt_init=False,
|
||||
# two stage
|
||||
two_stage_type="no", # ['no', 'standard', 'early', 'combine', 'enceachlayer', 'enclayer1']
|
||||
embed_init_tgt=False,
|
||||
# for text
|
||||
use_text_enhancer=False,
|
||||
use_fusion_layer=False,
|
||||
use_checkpoint=False,
|
||||
use_transformer_ckpt=False,
|
||||
use_text_cross_attention=False,
|
||||
text_dropout=0.1,
|
||||
fusion_dropout=0.1,
|
||||
fusion_droppath=0.0,
|
||||
):
|
||||
super().__init__()
|
||||
self.num_feature_levels = num_feature_levels
|
||||
self.num_encoder_layers = num_encoder_layers
|
||||
self.num_unicoder_layers = num_unicoder_layers
|
||||
self.num_decoder_layers = num_decoder_layers
|
||||
self.num_queries = num_queries
|
||||
assert query_dim == 4
|
||||
|
||||
# choose encoder layer type
|
||||
encoder_layer = DeformableTransformerEncoderLayer(
|
||||
d_model, dim_feedforward, dropout, activation, num_feature_levels, nhead, enc_n_points
|
||||
)
|
||||
|
||||
if use_text_enhancer:
|
||||
text_enhance_layer = TransformerEncoderLayer(
|
||||
d_model=d_model,
|
||||
nhead=nhead // 2,
|
||||
dim_feedforward=dim_feedforward // 2,
|
||||
dropout=text_dropout,
|
||||
)
|
||||
else:
|
||||
text_enhance_layer = None
|
||||
|
||||
if use_fusion_layer:
|
||||
feature_fusion_layer = BiAttentionBlock(
|
||||
v_dim=d_model,
|
||||
l_dim=d_model,
|
||||
embed_dim=dim_feedforward // 2,
|
||||
num_heads=nhead // 2,
|
||||
dropout=fusion_dropout,
|
||||
drop_path=fusion_droppath,
|
||||
)
|
||||
else:
|
||||
feature_fusion_layer = None
|
||||
|
||||
encoder_norm = nn.LayerNorm(d_model) if normalize_before else None
|
||||
assert encoder_norm is None
|
||||
self.encoder = TransformerEncoder(
|
||||
encoder_layer,
|
||||
num_encoder_layers,
|
||||
d_model=d_model,
|
||||
num_queries=num_queries,
|
||||
text_enhance_layer=text_enhance_layer,
|
||||
feature_fusion_layer=feature_fusion_layer,
|
||||
use_checkpoint=use_checkpoint,
|
||||
use_transformer_ckpt=use_transformer_ckpt,
|
||||
)
|
||||
|
||||
# choose decoder layer type
|
||||
decoder_layer = DeformableTransformerDecoderLayer(
|
||||
d_model,
|
||||
dim_feedforward,
|
||||
dropout,
|
||||
activation,
|
||||
num_feature_levels,
|
||||
nhead,
|
||||
dec_n_points,
|
||||
use_text_cross_attention=use_text_cross_attention,
|
||||
)
|
||||
|
||||
decoder_norm = nn.LayerNorm(d_model)
|
||||
self.decoder = TransformerDecoder(
|
||||
decoder_layer,
|
||||
num_decoder_layers,
|
||||
decoder_norm,
|
||||
return_intermediate=return_intermediate_dec,
|
||||
d_model=d_model,
|
||||
query_dim=query_dim,
|
||||
num_feature_levels=num_feature_levels,
|
||||
)
|
||||
|
||||
self.d_model = d_model
|
||||
self.nhead = nhead
|
||||
self.dec_layers = num_decoder_layers
|
||||
self.num_queries = num_queries # useful for single stage model only
|
||||
self.num_patterns = num_patterns
|
||||
if not isinstance(num_patterns, int):
|
||||
Warning("num_patterns should be int but {}".format(type(num_patterns)))
|
||||
self.num_patterns = 0
|
||||
|
||||
if num_feature_levels > 1:
|
||||
if self.num_encoder_layers > 0:
|
||||
self.level_embed = nn.Parameter(torch.Tensor(num_feature_levels, d_model))
|
||||
else:
|
||||
self.level_embed = None
|
||||
|
||||
self.learnable_tgt_init = learnable_tgt_init
|
||||
assert learnable_tgt_init, "why not learnable_tgt_init"
|
||||
self.embed_init_tgt = embed_init_tgt
|
||||
if (two_stage_type != "no" and embed_init_tgt) or (two_stage_type == "no"):
|
||||
self.tgt_embed = nn.Embedding(self.num_queries, d_model)
|
||||
nn.init.normal_(self.tgt_embed.weight.data)
|
||||
else:
|
||||
self.tgt_embed = None
|
||||
|
||||
# for two stage
|
||||
self.two_stage_type = two_stage_type
|
||||
assert two_stage_type in ["no", "standard"], "unknown param {} of two_stage_type".format(
|
||||
two_stage_type
|
||||
)
|
||||
if two_stage_type == "standard":
|
||||
# anchor selection at the output of encoder
|
||||
self.enc_output = nn.Linear(d_model, d_model)
|
||||
self.enc_output_norm = nn.LayerNorm(d_model)
|
||||
self.two_stage_wh_embedding = None
|
||||
|
||||
if two_stage_type == "no":
|
||||
self.init_ref_points(num_queries) # init self.refpoint_embed
|
||||
|
||||
self.enc_out_class_embed = None
|
||||
self.enc_out_bbox_embed = None
|
||||
|
||||
self._reset_parameters()
|
||||
|
||||
def _reset_parameters(self):
|
||||
for p in self.parameters():
|
||||
if p.dim() > 1:
|
||||
nn.init.xavier_uniform_(p)
|
||||
for m in self.modules():
|
||||
if isinstance(m, MSDeformAttn):
|
||||
m._reset_parameters()
|
||||
if self.num_feature_levels > 1 and self.level_embed is not None:
|
||||
nn.init.normal_(self.level_embed)
|
||||
|
||||
def get_valid_ratio(self, mask):
|
||||
_, H, W = mask.shape
|
||||
valid_H = torch.sum(~mask[:, :, 0], 1)
|
||||
valid_W = torch.sum(~mask[:, 0, :], 1)
|
||||
valid_ratio_h = valid_H.float() / H
|
||||
valid_ratio_w = valid_W.float() / W
|
||||
valid_ratio = torch.stack([valid_ratio_w, valid_ratio_h], -1)
|
||||
return valid_ratio
|
||||
|
||||
def init_ref_points(self, use_num_queries):
|
||||
self.refpoint_embed = nn.Embedding(use_num_queries, 4)
|
||||
|
||||
def forward(self, srcs, masks, refpoint_embed, pos_embeds, tgt, attn_mask=None, text_dict=None):
|
||||
"""
|
||||
Input:
|
||||
- srcs: List of multi features [bs, ci, hi, wi]
|
||||
- masks: List of multi masks [bs, hi, wi]
|
||||
- refpoint_embed: [bs, num_dn, 4]. None in infer
|
||||
- pos_embeds: List of multi pos embeds [bs, ci, hi, wi]
|
||||
- tgt: [bs, num_dn, d_model]. None in infer
|
||||
|
||||
"""
|
||||
# prepare input for encoder
|
||||
src_flatten = []
|
||||
mask_flatten = []
|
||||
lvl_pos_embed_flatten = []
|
||||
spatial_shapes = []
|
||||
for lvl, (src, mask, pos_embed) in enumerate(zip(srcs, masks, pos_embeds)):
|
||||
bs, c, h, w = src.shape
|
||||
spatial_shape = (h, w)
|
||||
spatial_shapes.append(spatial_shape)
|
||||
|
||||
src = src.flatten(2).transpose(1, 2) # bs, hw, c
|
||||
mask = mask.flatten(1) # bs, hw
|
||||
pos_embed = pos_embed.flatten(2).transpose(1, 2) # bs, hw, c
|
||||
if self.num_feature_levels > 1 and self.level_embed is not None:
|
||||
lvl_pos_embed = pos_embed + self.level_embed[lvl].view(1, 1, -1)
|
||||
else:
|
||||
lvl_pos_embed = pos_embed
|
||||
lvl_pos_embed_flatten.append(lvl_pos_embed)
|
||||
src_flatten.append(src)
|
||||
mask_flatten.append(mask)
|
||||
src_flatten = torch.cat(src_flatten, 1) # bs, \sum{hxw}, c
|
||||
mask_flatten = torch.cat(mask_flatten, 1) # bs, \sum{hxw}
|
||||
lvl_pos_embed_flatten = torch.cat(lvl_pos_embed_flatten, 1) # bs, \sum{hxw}, c
|
||||
spatial_shapes = torch.as_tensor(
|
||||
spatial_shapes, dtype=torch.long, device=src_flatten.device
|
||||
)
|
||||
level_start_index = torch.cat(
|
||||
(spatial_shapes.new_zeros((1,)), spatial_shapes.prod(1).cumsum(0)[:-1])
|
||||
)
|
||||
valid_ratios = torch.stack([self.get_valid_ratio(m) for m in masks], 1)
|
||||
|
||||
# two stage
|
||||
enc_topk_proposals = enc_refpoint_embed = None
|
||||
|
||||
#########################################################
|
||||
# Begin Encoder
|
||||
#########################################################
|
||||
memory, memory_text = self.encoder(
|
||||
src_flatten,
|
||||
pos=lvl_pos_embed_flatten,
|
||||
level_start_index=level_start_index,
|
||||
spatial_shapes=spatial_shapes,
|
||||
valid_ratios=valid_ratios,
|
||||
key_padding_mask=mask_flatten,
|
||||
memory_text=text_dict["encoded_text"],
|
||||
text_attention_mask=~text_dict["text_token_mask"],
|
||||
# we ~ the mask . False means use the token; True means pad the token
|
||||
position_ids=text_dict["position_ids"],
|
||||
text_self_attention_masks=text_dict["text_self_attention_masks"],
|
||||
)
|
||||
#########################################################
|
||||
# End Encoder
|
||||
# - memory: bs, \sum{hw}, c
|
||||
# - mask_flatten: bs, \sum{hw}
|
||||
# - lvl_pos_embed_flatten: bs, \sum{hw}, c
|
||||
# - enc_intermediate_output: None or (nenc+1, bs, nq, c) or (nenc, bs, nq, c)
|
||||
# - enc_intermediate_refpoints: None or (nenc+1, bs, nq, c) or (nenc, bs, nq, c)
|
||||
#########################################################
|
||||
text_dict["encoded_text"] = memory_text
|
||||
# if os.environ.get("SHILONG_AMP_INFNAN_DEBUG") == '1':
|
||||
# if memory.isnan().any() | memory.isinf().any():
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
if self.two_stage_type == "standard":
|
||||
output_memory, output_proposals = gen_encoder_output_proposals(
|
||||
memory, mask_flatten, spatial_shapes
|
||||
)
|
||||
output_memory = self.enc_output_norm(self.enc_output(output_memory))
|
||||
|
||||
if text_dict is not None:
|
||||
enc_outputs_class_unselected = self.enc_out_class_embed(output_memory, text_dict)
|
||||
else:
|
||||
enc_outputs_class_unselected = self.enc_out_class_embed(output_memory)
|
||||
|
||||
topk_logits = enc_outputs_class_unselected.max(-1)[0]
|
||||
enc_outputs_coord_unselected = (
|
||||
self.enc_out_bbox_embed(output_memory) + output_proposals
|
||||
) # (bs, \sum{hw}, 4) unsigmoid
|
||||
topk = self.num_queries
|
||||
|
||||
topk_proposals = torch.topk(topk_logits, topk, dim=1)[1] # bs, nq
|
||||
|
||||
# gather boxes
|
||||
refpoint_embed_undetach = torch.gather(
|
||||
enc_outputs_coord_unselected, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, 4)
|
||||
) # unsigmoid
|
||||
refpoint_embed_ = refpoint_embed_undetach.detach()
|
||||
init_box_proposal = torch.gather(
|
||||
output_proposals, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, 4)
|
||||
).sigmoid() # sigmoid
|
||||
|
||||
# gather tgt
|
||||
tgt_undetach = torch.gather(
|
||||
output_memory, 1, topk_proposals.unsqueeze(-1).repeat(1, 1, self.d_model)
|
||||
)
|
||||
if self.embed_init_tgt:
|
||||
tgt_ = (
|
||||
self.tgt_embed.weight[:, None, :].repeat(1, bs, 1).transpose(0, 1)
|
||||
) # nq, bs, d_model
|
||||
else:
|
||||
tgt_ = tgt_undetach.detach()
|
||||
|
||||
if refpoint_embed is not None:
|
||||
refpoint_embed = torch.cat([refpoint_embed, refpoint_embed_], dim=1)
|
||||
tgt = torch.cat([tgt, tgt_], dim=1)
|
||||
else:
|
||||
refpoint_embed, tgt = refpoint_embed_, tgt_
|
||||
|
||||
elif self.two_stage_type == "no":
|
||||
tgt_ = (
|
||||
self.tgt_embed.weight[:, None, :].repeat(1, bs, 1).transpose(0, 1)
|
||||
) # nq, bs, d_model
|
||||
refpoint_embed_ = (
|
||||
self.refpoint_embed.weight[:, None, :].repeat(1, bs, 1).transpose(0, 1)
|
||||
) # nq, bs, 4
|
||||
|
||||
if refpoint_embed is not None:
|
||||
refpoint_embed = torch.cat([refpoint_embed, refpoint_embed_], dim=1)
|
||||
tgt = torch.cat([tgt, tgt_], dim=1)
|
||||
else:
|
||||
refpoint_embed, tgt = refpoint_embed_, tgt_
|
||||
|
||||
if self.num_patterns > 0:
|
||||
tgt_embed = tgt.repeat(1, self.num_patterns, 1)
|
||||
refpoint_embed = refpoint_embed.repeat(1, self.num_patterns, 1)
|
||||
tgt_pat = self.patterns.weight[None, :, :].repeat_interleave(
|
||||
self.num_queries, 1
|
||||
) # 1, n_q*n_pat, d_model
|
||||
tgt = tgt_embed + tgt_pat
|
||||
|
||||
init_box_proposal = refpoint_embed_.sigmoid()
|
||||
|
||||
else:
|
||||
raise NotImplementedError("unknown two_stage_type {}".format(self.two_stage_type))
|
||||
#########################################################
|
||||
# End preparing tgt
|
||||
# - tgt: bs, NQ, d_model
|
||||
# - refpoint_embed(unsigmoid): bs, NQ, d_model
|
||||
#########################################################
|
||||
|
||||
#########################################################
|
||||
# Begin Decoder
|
||||
#########################################################
|
||||
hs, references = self.decoder(
|
||||
tgt=tgt.transpose(0, 1),
|
||||
memory=memory.transpose(0, 1),
|
||||
memory_key_padding_mask=mask_flatten,
|
||||
pos=lvl_pos_embed_flatten.transpose(0, 1),
|
||||
refpoints_unsigmoid=refpoint_embed.transpose(0, 1),
|
||||
level_start_index=level_start_index,
|
||||
spatial_shapes=spatial_shapes,
|
||||
valid_ratios=valid_ratios,
|
||||
tgt_mask=attn_mask,
|
||||
memory_text=text_dict["encoded_text"],
|
||||
text_attention_mask=~text_dict["text_token_mask"],
|
||||
# we ~ the mask . False means use the token; True means pad the token
|
||||
)
|
||||
#########################################################
|
||||
# End Decoder
|
||||
# hs: n_dec, bs, nq, d_model
|
||||
# references: n_dec+1, bs, nq, query_dim
|
||||
#########################################################
|
||||
|
||||
#########################################################
|
||||
# Begin postprocess
|
||||
#########################################################
|
||||
if self.two_stage_type == "standard":
|
||||
hs_enc = tgt_undetach.unsqueeze(0)
|
||||
ref_enc = refpoint_embed_undetach.sigmoid().unsqueeze(0)
|
||||
else:
|
||||
hs_enc = ref_enc = None
|
||||
#########################################################
|
||||
# End postprocess
|
||||
# hs_enc: (n_enc+1, bs, nq, d_model) or (1, bs, nq, d_model) or (n_enc, bs, nq, d_model) or None
|
||||
# ref_enc: (n_enc+1, bs, nq, query_dim) or (1, bs, nq, query_dim) or (n_enc, bs, nq, d_model) or None
|
||||
#########################################################
|
||||
|
||||
return hs, references, hs_enc, ref_enc, init_box_proposal
|
||||
# hs: (n_dec, bs, nq, d_model)
|
||||
# references: sigmoid coordinates. (n_dec+1, bs, bq, 4)
|
||||
# hs_enc: (n_enc+1, bs, nq, d_model) or (1, bs, nq, d_model) or None
|
||||
# ref_enc: sigmoid coordinates. \
|
||||
# (n_enc+1, bs, nq, query_dim) or (1, bs, nq, query_dim) or None
|
||||
|
||||
|
||||
class TransformerEncoder(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
encoder_layer,
|
||||
num_layers,
|
||||
d_model=256,
|
||||
num_queries=300,
|
||||
enc_layer_share=False,
|
||||
text_enhance_layer=None,
|
||||
feature_fusion_layer=None,
|
||||
use_checkpoint=False,
|
||||
use_transformer_ckpt=False,
|
||||
):
|
||||
"""_summary_
|
||||
|
||||
Args:
|
||||
encoder_layer (_type_): _description_
|
||||
num_layers (_type_): _description_
|
||||
norm (_type_, optional): _description_. Defaults to None.
|
||||
d_model (int, optional): _description_. Defaults to 256.
|
||||
num_queries (int, optional): _description_. Defaults to 300.
|
||||
enc_layer_share (bool, optional): _description_. Defaults to False.
|
||||
|
||||
"""
|
||||
super().__init__()
|
||||
# prepare layers
|
||||
self.layers = []
|
||||
self.text_layers = []
|
||||
self.fusion_layers = []
|
||||
if num_layers > 0:
|
||||
self.layers = _get_clones(encoder_layer, num_layers, layer_share=enc_layer_share)
|
||||
|
||||
if text_enhance_layer is not None:
|
||||
self.text_layers = _get_clones(
|
||||
text_enhance_layer, num_layers, layer_share=enc_layer_share
|
||||
)
|
||||
if feature_fusion_layer is not None:
|
||||
self.fusion_layers = _get_clones(
|
||||
feature_fusion_layer, num_layers, layer_share=enc_layer_share
|
||||
)
|
||||
else:
|
||||
self.layers = []
|
||||
del encoder_layer
|
||||
|
||||
if text_enhance_layer is not None:
|
||||
self.text_layers = []
|
||||
del text_enhance_layer
|
||||
if feature_fusion_layer is not None:
|
||||
self.fusion_layers = []
|
||||
del feature_fusion_layer
|
||||
|
||||
self.query_scale = None
|
||||
self.num_queries = num_queries
|
||||
self.num_layers = num_layers
|
||||
self.d_model = d_model
|
||||
|
||||
self.use_checkpoint = use_checkpoint
|
||||
self.use_transformer_ckpt = use_transformer_ckpt
|
||||
|
||||
@staticmethod
|
||||
def get_reference_points(spatial_shapes, valid_ratios, device):
|
||||
reference_points_list = []
|
||||
for lvl, (H_, W_) in enumerate(spatial_shapes):
|
||||
|
||||
ref_y, ref_x = torch.meshgrid(
|
||||
torch.linspace(0.5, H_ - 0.5, H_, dtype=torch.float32, device=device),
|
||||
torch.linspace(0.5, W_ - 0.5, W_, dtype=torch.float32, device=device),
|
||||
)
|
||||
ref_y = ref_y.reshape(-1)[None] / (valid_ratios[:, None, lvl, 1] * H_)
|
||||
ref_x = ref_x.reshape(-1)[None] / (valid_ratios[:, None, lvl, 0] * W_)
|
||||
ref = torch.stack((ref_x, ref_y), -1)
|
||||
reference_points_list.append(ref)
|
||||
reference_points = torch.cat(reference_points_list, 1)
|
||||
reference_points = reference_points[:, :, None] * valid_ratios[:, None]
|
||||
return reference_points
|
||||
|
||||
def forward(
|
||||
self,
|
||||
# for images
|
||||
src: Tensor,
|
||||
pos: Tensor,
|
||||
spatial_shapes: Tensor,
|
||||
level_start_index: Tensor,
|
||||
valid_ratios: Tensor,
|
||||
key_padding_mask: Tensor,
|
||||
# for texts
|
||||
memory_text: Tensor = None,
|
||||
text_attention_mask: Tensor = None,
|
||||
pos_text: Tensor = None,
|
||||
text_self_attention_masks: Tensor = None,
|
||||
position_ids: Tensor = None,
|
||||
):
|
||||
"""
|
||||
Input:
|
||||
- src: [bs, sum(hi*wi), 256]
|
||||
- pos: pos embed for src. [bs, sum(hi*wi), 256]
|
||||
- spatial_shapes: h,w of each level [num_level, 2]
|
||||
- level_start_index: [num_level] start point of level in sum(hi*wi).
|
||||
- valid_ratios: [bs, num_level, 2]
|
||||
- key_padding_mask: [bs, sum(hi*wi)]
|
||||
|
||||
- memory_text: bs, n_text, 256
|
||||
- text_attention_mask: bs, n_text
|
||||
False for no padding; True for padding
|
||||
- pos_text: bs, n_text, 256
|
||||
|
||||
- position_ids: bs, n_text
|
||||
Intermedia:
|
||||
- reference_points: [bs, sum(hi*wi), num_level, 2]
|
||||
Outpus:
|
||||
- output: [bs, sum(hi*wi), 256]
|
||||
"""
|
||||
|
||||
output = src
|
||||
|
||||
# preparation and reshape
|
||||
if self.num_layers > 0:
|
||||
reference_points = self.get_reference_points(
|
||||
spatial_shapes, valid_ratios, device=src.device
|
||||
)
|
||||
|
||||
if self.text_layers:
|
||||
# generate pos_text
|
||||
bs, n_text, text_dim = memory_text.shape
|
||||
if pos_text is None and position_ids is None:
|
||||
pos_text = (
|
||||
torch.arange(n_text, device=memory_text.device)
|
||||
.float()
|
||||
.unsqueeze(0)
|
||||
.unsqueeze(-1)
|
||||
.repeat(bs, 1, 1)
|
||||
)
|
||||
pos_text = get_sine_pos_embed(pos_text, num_pos_feats=256, exchange_xy=False)
|
||||
if position_ids is not None:
|
||||
pos_text = get_sine_pos_embed(
|
||||
position_ids[..., None], num_pos_feats=256, exchange_xy=False
|
||||
)
|
||||
|
||||
# main process
|
||||
for layer_id, layer in enumerate(self.layers):
|
||||
# if output.isnan().any() or memory_text.isnan().any():
|
||||
# if os.environ.get('IPDB_SHILONG_DEBUG', None) == 'INFO':
|
||||
# import ipdb; ipdb.set_trace()
|
||||
if self.fusion_layers:
|
||||
if self.use_checkpoint:
|
||||
output, memory_text = checkpoint.checkpoint(
|
||||
self.fusion_layers[layer_id],
|
||||
output,
|
||||
memory_text,
|
||||
key_padding_mask,
|
||||
text_attention_mask,
|
||||
)
|
||||
else:
|
||||
output, memory_text = self.fusion_layers[layer_id](
|
||||
v=output,
|
||||
l=memory_text,
|
||||
attention_mask_v=key_padding_mask,
|
||||
attention_mask_l=text_attention_mask,
|
||||
)
|
||||
|
||||
if self.text_layers:
|
||||
memory_text = self.text_layers[layer_id](
|
||||
src=memory_text.transpose(0, 1),
|
||||
src_mask=~text_self_attention_masks, # note we use ~ for mask here
|
||||
src_key_padding_mask=text_attention_mask,
|
||||
pos=(pos_text.transpose(0, 1) if pos_text is not None else None),
|
||||
).transpose(0, 1)
|
||||
|
||||
# main process
|
||||
if self.use_transformer_ckpt:
|
||||
output = checkpoint.checkpoint(
|
||||
layer,
|
||||
output,
|
||||
pos,
|
||||
reference_points,
|
||||
spatial_shapes,
|
||||
level_start_index,
|
||||
key_padding_mask,
|
||||
)
|
||||
else:
|
||||
output = layer(
|
||||
src=output,
|
||||
pos=pos,
|
||||
reference_points=reference_points,
|
||||
spatial_shapes=spatial_shapes,
|
||||
level_start_index=level_start_index,
|
||||
key_padding_mask=key_padding_mask,
|
||||
)
|
||||
|
||||
return output, memory_text
|
||||
|
||||
|
||||
class TransformerDecoder(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
decoder_layer,
|
||||
num_layers,
|
||||
norm=None,
|
||||
return_intermediate=False,
|
||||
d_model=256,
|
||||
query_dim=4,
|
||||
num_feature_levels=1,
|
||||
):
|
||||
super().__init__()
|
||||
if num_layers > 0:
|
||||
self.layers = _get_clones(decoder_layer, num_layers)
|
||||
else:
|
||||
self.layers = []
|
||||
self.num_layers = num_layers
|
||||
self.norm = norm
|
||||
self.return_intermediate = return_intermediate
|
||||
assert return_intermediate, "support return_intermediate only"
|
||||
self.query_dim = query_dim
|
||||
assert query_dim in [2, 4], "query_dim should be 2/4 but {}".format(query_dim)
|
||||
self.num_feature_levels = num_feature_levels
|
||||
|
||||
self.ref_point_head = MLP(query_dim // 2 * d_model, d_model, d_model, 2)
|
||||
self.query_pos_sine_scale = None
|
||||
|
||||
self.query_scale = None
|
||||
self.bbox_embed = None
|
||||
self.class_embed = None
|
||||
|
||||
self.d_model = d_model
|
||||
|
||||
self.ref_anchor_head = None
|
||||
|
||||
def forward(
|
||||
self,
|
||||
tgt,
|
||||
memory,
|
||||
tgt_mask: Optional[Tensor] = None,
|
||||
memory_mask: Optional[Tensor] = None,
|
||||
tgt_key_padding_mask: Optional[Tensor] = None,
|
||||
memory_key_padding_mask: Optional[Tensor] = None,
|
||||
pos: Optional[Tensor] = None,
|
||||
refpoints_unsigmoid: Optional[Tensor] = None, # num_queries, bs, 2
|
||||
# for memory
|
||||
level_start_index: Optional[Tensor] = None, # num_levels
|
||||
spatial_shapes: Optional[Tensor] = None, # bs, num_levels, 2
|
||||
valid_ratios: Optional[Tensor] = None,
|
||||
# for text
|
||||
memory_text: Optional[Tensor] = None,
|
||||
text_attention_mask: Optional[Tensor] = None,
|
||||
):
|
||||
"""
|
||||
Input:
|
||||
- tgt: nq, bs, d_model
|
||||
- memory: hw, bs, d_model
|
||||
- pos: hw, bs, d_model
|
||||
- refpoints_unsigmoid: nq, bs, 2/4
|
||||
- valid_ratios/spatial_shapes: bs, nlevel, 2
|
||||
"""
|
||||
output = tgt
|
||||
|
||||
intermediate = []
|
||||
reference_points = refpoints_unsigmoid.sigmoid()
|
||||
ref_points = [reference_points]
|
||||
|
||||
for layer_id, layer in enumerate(self.layers):
|
||||
|
||||
if reference_points.shape[-1] == 4:
|
||||
reference_points_input = (
|
||||
reference_points[:, :, None]
|
||||
* torch.cat([valid_ratios, valid_ratios], -1)[None, :]
|
||||
) # nq, bs, nlevel, 4
|
||||
else:
|
||||
assert reference_points.shape[-1] == 2
|
||||
reference_points_input = reference_points[:, :, None] * valid_ratios[None, :]
|
||||
query_sine_embed = gen_sineembed_for_position(
|
||||
reference_points_input[:, :, 0, :]
|
||||
) # nq, bs, 256*2
|
||||
|
||||
# conditional query
|
||||
raw_query_pos = self.ref_point_head(query_sine_embed) # nq, bs, 256
|
||||
pos_scale = self.query_scale(output) if self.query_scale is not None else 1
|
||||
query_pos = pos_scale * raw_query_pos
|
||||
# if os.environ.get("SHILONG_AMP_INFNAN_DEBUG") == '1':
|
||||
# if query_pos.isnan().any() | query_pos.isinf().any():
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
# main process
|
||||
output = layer(
|
||||
tgt=output,
|
||||
tgt_query_pos=query_pos,
|
||||
tgt_query_sine_embed=query_sine_embed,
|
||||
tgt_key_padding_mask=tgt_key_padding_mask,
|
||||
tgt_reference_points=reference_points_input,
|
||||
memory_text=memory_text,
|
||||
text_attention_mask=text_attention_mask,
|
||||
memory=memory,
|
||||
memory_key_padding_mask=memory_key_padding_mask,
|
||||
memory_level_start_index=level_start_index,
|
||||
memory_spatial_shapes=spatial_shapes,
|
||||
memory_pos=pos,
|
||||
self_attn_mask=tgt_mask,
|
||||
cross_attn_mask=memory_mask,
|
||||
)
|
||||
if output.isnan().any() | output.isinf().any():
|
||||
print(f"output layer_id {layer_id} is nan")
|
||||
try:
|
||||
num_nan = output.isnan().sum().item()
|
||||
num_inf = output.isinf().sum().item()
|
||||
print(f"num_nan {num_nan}, num_inf {num_inf}")
|
||||
except Exception as e:
|
||||
print(e)
|
||||
# if os.environ.get("SHILONG_AMP_INFNAN_DEBUG") == '1':
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
# iter update
|
||||
if self.bbox_embed is not None:
|
||||
# box_holder = self.bbox_embed(output)
|
||||
# box_holder[..., :self.query_dim] += inverse_sigmoid(reference_points)
|
||||
# new_reference_points = box_holder[..., :self.query_dim].sigmoid()
|
||||
|
||||
reference_before_sigmoid = inverse_sigmoid(reference_points)
|
||||
delta_unsig = self.bbox_embed[layer_id](output)
|
||||
outputs_unsig = delta_unsig + reference_before_sigmoid
|
||||
new_reference_points = outputs_unsig.sigmoid()
|
||||
|
||||
reference_points = new_reference_points.detach()
|
||||
# if layer_id != self.num_layers - 1:
|
||||
ref_points.append(new_reference_points)
|
||||
|
||||
intermediate.append(self.norm(output))
|
||||
|
||||
return [
|
||||
[itm_out.transpose(0, 1) for itm_out in intermediate],
|
||||
[itm_refpoint.transpose(0, 1) for itm_refpoint in ref_points],
|
||||
]
|
||||
|
||||
|
||||
class DeformableTransformerEncoderLayer(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
d_model=256,
|
||||
d_ffn=1024,
|
||||
dropout=0.1,
|
||||
activation="relu",
|
||||
n_levels=4,
|
||||
n_heads=8,
|
||||
n_points=4,
|
||||
):
|
||||
super().__init__()
|
||||
|
||||
# self attention
|
||||
self.self_attn = MSDeformAttn(
|
||||
embed_dim=d_model,
|
||||
num_levels=n_levels,
|
||||
num_heads=n_heads,
|
||||
num_points=n_points,
|
||||
batch_first=True,
|
||||
)
|
||||
self.dropout1 = nn.Dropout(dropout)
|
||||
self.norm1 = nn.LayerNorm(d_model)
|
||||
|
||||
# ffn
|
||||
self.linear1 = nn.Linear(d_model, d_ffn)
|
||||
self.activation = _get_activation_fn(activation, d_model=d_ffn)
|
||||
self.dropout2 = nn.Dropout(dropout)
|
||||
self.linear2 = nn.Linear(d_ffn, d_model)
|
||||
self.dropout3 = nn.Dropout(dropout)
|
||||
self.norm2 = nn.LayerNorm(d_model)
|
||||
|
||||
@staticmethod
|
||||
def with_pos_embed(tensor, pos):
|
||||
return tensor if pos is None else tensor + pos
|
||||
|
||||
def forward_ffn(self, src):
|
||||
src2 = self.linear2(self.dropout2(self.activation(self.linear1(src))))
|
||||
src = src + self.dropout3(src2)
|
||||
src = self.norm2(src)
|
||||
return src
|
||||
|
||||
def forward(
|
||||
self, src, pos, reference_points, spatial_shapes, level_start_index, key_padding_mask=None
|
||||
):
|
||||
# self attention
|
||||
# import ipdb; ipdb.set_trace()
|
||||
src2 = self.self_attn(
|
||||
query=self.with_pos_embed(src, pos),
|
||||
reference_points=reference_points,
|
||||
value=src,
|
||||
spatial_shapes=spatial_shapes,
|
||||
level_start_index=level_start_index,
|
||||
key_padding_mask=key_padding_mask,
|
||||
)
|
||||
src = src + self.dropout1(src2)
|
||||
src = self.norm1(src)
|
||||
|
||||
# ffn
|
||||
src = self.forward_ffn(src)
|
||||
|
||||
return src
|
||||
|
||||
|
||||
class DeformableTransformerDecoderLayer(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
d_model=256,
|
||||
d_ffn=1024,
|
||||
dropout=0.1,
|
||||
activation="relu",
|
||||
n_levels=4,
|
||||
n_heads=8,
|
||||
n_points=4,
|
||||
use_text_feat_guide=False,
|
||||
use_text_cross_attention=False,
|
||||
):
|
||||
super().__init__()
|
||||
|
||||
# cross attention
|
||||
self.cross_attn = MSDeformAttn(
|
||||
embed_dim=d_model,
|
||||
num_levels=n_levels,
|
||||
num_heads=n_heads,
|
||||
num_points=n_points,
|
||||
batch_first=True,
|
||||
)
|
||||
self.dropout1 = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
||||
self.norm1 = nn.LayerNorm(d_model)
|
||||
|
||||
# cross attention text
|
||||
if use_text_cross_attention:
|
||||
self.ca_text = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
|
||||
self.catext_dropout = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
||||
self.catext_norm = nn.LayerNorm(d_model)
|
||||
|
||||
# self attention
|
||||
self.self_attn = nn.MultiheadAttention(d_model, n_heads, dropout=dropout)
|
||||
self.dropout2 = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
||||
self.norm2 = nn.LayerNorm(d_model)
|
||||
|
||||
# ffn
|
||||
self.linear1 = nn.Linear(d_model, d_ffn)
|
||||
self.activation = _get_activation_fn(activation, d_model=d_ffn, batch_dim=1)
|
||||
self.dropout3 = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
||||
self.linear2 = nn.Linear(d_ffn, d_model)
|
||||
self.dropout4 = nn.Dropout(dropout) if dropout > 0 else nn.Identity()
|
||||
self.norm3 = nn.LayerNorm(d_model)
|
||||
|
||||
self.key_aware_proj = None
|
||||
self.use_text_feat_guide = use_text_feat_guide
|
||||
assert not use_text_feat_guide
|
||||
self.use_text_cross_attention = use_text_cross_attention
|
||||
|
||||
def rm_self_attn_modules(self):
|
||||
self.self_attn = None
|
||||
self.dropout2 = None
|
||||
self.norm2 = None
|
||||
|
||||
@staticmethod
|
||||
def with_pos_embed(tensor, pos):
|
||||
return tensor if pos is None else tensor + pos
|
||||
|
||||
def forward_ffn(self, tgt):
|
||||
with torch.cuda.amp.autocast(enabled=False):
|
||||
tgt2 = self.linear2(self.dropout3(self.activation(self.linear1(tgt))))
|
||||
tgt = tgt + self.dropout4(tgt2)
|
||||
tgt = self.norm3(tgt)
|
||||
return tgt
|
||||
|
||||
def forward(
|
||||
self,
|
||||
# for tgt
|
||||
tgt: Optional[Tensor], # nq, bs, d_model
|
||||
tgt_query_pos: Optional[Tensor] = None, # pos for query. MLP(Sine(pos))
|
||||
tgt_query_sine_embed: Optional[Tensor] = None, # pos for query. Sine(pos)
|
||||
tgt_key_padding_mask: Optional[Tensor] = None,
|
||||
tgt_reference_points: Optional[Tensor] = None, # nq, bs, 4
|
||||
memory_text: Optional[Tensor] = None, # bs, num_token, d_model
|
||||
text_attention_mask: Optional[Tensor] = None, # bs, num_token
|
||||
# for memory
|
||||
memory: Optional[Tensor] = None, # hw, bs, d_model
|
||||
memory_key_padding_mask: Optional[Tensor] = None,
|
||||
memory_level_start_index: Optional[Tensor] = None, # num_levels
|
||||
memory_spatial_shapes: Optional[Tensor] = None, # bs, num_levels, 2
|
||||
memory_pos: Optional[Tensor] = None, # pos for memory
|
||||
# sa
|
||||
self_attn_mask: Optional[Tensor] = None, # mask used for self-attention
|
||||
cross_attn_mask: Optional[Tensor] = None, # mask used for cross-attention
|
||||
):
|
||||
"""
|
||||
Input:
|
||||
- tgt/tgt_query_pos: nq, bs, d_model
|
||||
-
|
||||
"""
|
||||
assert cross_attn_mask is None
|
||||
|
||||
# self attention
|
||||
if self.self_attn is not None:
|
||||
# import ipdb; ipdb.set_trace()
|
||||
q = k = self.with_pos_embed(tgt, tgt_query_pos)
|
||||
tgt2 = self.self_attn(q, k, tgt, attn_mask=self_attn_mask)[0]
|
||||
tgt = tgt + self.dropout2(tgt2)
|
||||
tgt = self.norm2(tgt)
|
||||
|
||||
if self.use_text_cross_attention:
|
||||
tgt2 = self.ca_text(
|
||||
self.with_pos_embed(tgt, tgt_query_pos),
|
||||
memory_text.transpose(0, 1),
|
||||
memory_text.transpose(0, 1),
|
||||
key_padding_mask=text_attention_mask,
|
||||
)[0]
|
||||
tgt = tgt + self.catext_dropout(tgt2)
|
||||
tgt = self.catext_norm(tgt)
|
||||
|
||||
tgt2 = self.cross_attn(
|
||||
query=self.with_pos_embed(tgt, tgt_query_pos).transpose(0, 1),
|
||||
reference_points=tgt_reference_points.transpose(0, 1).contiguous(),
|
||||
value=memory.transpose(0, 1),
|
||||
spatial_shapes=memory_spatial_shapes,
|
||||
level_start_index=memory_level_start_index,
|
||||
key_padding_mask=memory_key_padding_mask,
|
||||
).transpose(0, 1)
|
||||
tgt = tgt + self.dropout1(tgt2)
|
||||
tgt = self.norm1(tgt)
|
||||
|
||||
# ffn
|
||||
tgt = self.forward_ffn(tgt)
|
||||
|
||||
return tgt
|
||||
|
||||
|
||||
def build_transformer(args):
|
||||
return Transformer(
|
||||
d_model=args.hidden_dim,
|
||||
dropout=args.dropout,
|
||||
nhead=args.nheads,
|
||||
num_queries=args.num_queries,
|
||||
dim_feedforward=args.dim_feedforward,
|
||||
num_encoder_layers=args.enc_layers,
|
||||
num_decoder_layers=args.dec_layers,
|
||||
normalize_before=args.pre_norm,
|
||||
return_intermediate_dec=True,
|
||||
query_dim=args.query_dim,
|
||||
activation=args.transformer_activation,
|
||||
num_patterns=args.num_patterns,
|
||||
num_feature_levels=args.num_feature_levels,
|
||||
enc_n_points=args.enc_n_points,
|
||||
dec_n_points=args.dec_n_points,
|
||||
learnable_tgt_init=True,
|
||||
# two stage
|
||||
two_stage_type=args.two_stage_type, # ['no', 'standard', 'early']
|
||||
embed_init_tgt=args.embed_init_tgt,
|
||||
use_text_enhancer=args.use_text_enhancer,
|
||||
use_fusion_layer=args.use_fusion_layer,
|
||||
use_checkpoint=args.use_checkpoint,
|
||||
use_transformer_ckpt=args.use_transformer_ckpt,
|
||||
use_text_cross_attention=args.use_text_cross_attention,
|
||||
text_dropout=args.text_dropout,
|
||||
fusion_dropout=args.fusion_dropout,
|
||||
fusion_droppath=args.fusion_droppath,
|
||||
)
|
||||
|
|
@ -0,0 +1,123 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Copyright (c) Aishwarya Kamath & Nicolas Carion. Licensed under the Apache License 2.0. All Rights Reserved
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
"""
|
||||
DETR Transformer class.
|
||||
|
||||
Copy-paste from torch.nn.Transformer with modifications:
|
||||
* positional encodings are passed in MHattention
|
||||
* extra LN at the end of encoder is removed
|
||||
* decoder returns a stack of activations from all decoding layers
|
||||
"""
|
||||
from typing import Optional
|
||||
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import Tensor, nn
|
||||
|
||||
from .utils import (
|
||||
MLP,
|
||||
_get_activation_fn,
|
||||
_get_clones,
|
||||
gen_encoder_output_proposals,
|
||||
gen_sineembed_for_position,
|
||||
sigmoid_focal_loss,
|
||||
)
|
||||
|
||||
|
||||
class TextTransformer(nn.Module):
|
||||
def __init__(self, num_layers, d_model=256, nheads=8, dim_feedforward=2048, dropout=0.1):
|
||||
super().__init__()
|
||||
self.num_layers = num_layers
|
||||
self.d_model = d_model
|
||||
self.nheads = nheads
|
||||
self.dim_feedforward = dim_feedforward
|
||||
self.norm = None
|
||||
|
||||
single_encoder_layer = TransformerEncoderLayer(
|
||||
d_model=d_model, nhead=nheads, dim_feedforward=dim_feedforward, dropout=dropout
|
||||
)
|
||||
self.layers = _get_clones(single_encoder_layer, num_layers)
|
||||
|
||||
def forward(self, memory_text: torch.Tensor, text_attention_mask: torch.Tensor):
|
||||
"""
|
||||
|
||||
Args:
|
||||
text_attention_mask: bs, num_token
|
||||
memory_text: bs, num_token, d_model
|
||||
|
||||
Raises:
|
||||
RuntimeError: _description_
|
||||
|
||||
Returns:
|
||||
output: bs, num_token, d_model
|
||||
"""
|
||||
|
||||
output = memory_text.transpose(0, 1)
|
||||
|
||||
for layer in self.layers:
|
||||
output = layer(output, src_key_padding_mask=text_attention_mask)
|
||||
|
||||
if self.norm is not None:
|
||||
output = self.norm(output)
|
||||
|
||||
return output.transpose(0, 1)
|
||||
|
||||
|
||||
class TransformerEncoderLayer(nn.Module):
|
||||
def __init__(
|
||||
self,
|
||||
d_model,
|
||||
nhead,
|
||||
dim_feedforward=2048,
|
||||
dropout=0.1,
|
||||
activation="relu",
|
||||
normalize_before=False,
|
||||
):
|
||||
super().__init__()
|
||||
self.self_attn = nn.MultiheadAttention(d_model, nhead, dropout=dropout)
|
||||
# Implementation of Feedforward model
|
||||
self.linear1 = nn.Linear(d_model, dim_feedforward)
|
||||
self.dropout = nn.Dropout(dropout)
|
||||
self.linear2 = nn.Linear(dim_feedforward, d_model)
|
||||
|
||||
self.norm1 = nn.LayerNorm(d_model)
|
||||
self.norm2 = nn.LayerNorm(d_model)
|
||||
self.dropout1 = nn.Dropout(dropout)
|
||||
self.dropout2 = nn.Dropout(dropout)
|
||||
|
||||
self.activation = _get_activation_fn(activation)
|
||||
self.normalize_before = normalize_before
|
||||
self.nhead = nhead
|
||||
|
||||
def with_pos_embed(self, tensor, pos: Optional[Tensor]):
|
||||
return tensor if pos is None else tensor + pos
|
||||
|
||||
def forward(
|
||||
self,
|
||||
src,
|
||||
src_mask: Optional[Tensor] = None,
|
||||
src_key_padding_mask: Optional[Tensor] = None,
|
||||
pos: Optional[Tensor] = None,
|
||||
):
|
||||
# repeat attn mask
|
||||
if src_mask.dim() == 3 and src_mask.shape[0] == src.shape[1]:
|
||||
# bs, num_q, num_k
|
||||
src_mask = src_mask.repeat(self.nhead, 1, 1)
|
||||
|
||||
q = k = self.with_pos_embed(src, pos)
|
||||
|
||||
src2 = self.self_attn(q, k, value=src, attn_mask=src_mask)[0]
|
||||
|
||||
# src2 = self.self_attn(q, k, value=src, attn_mask=src_mask, key_padding_mask=src_key_padding_mask)[0]
|
||||
src = src + self.dropout1(src2)
|
||||
src = self.norm1(src)
|
||||
src2 = self.linear2(self.dropout(self.activation(self.linear1(src))))
|
||||
src = src + self.dropout2(src2)
|
||||
src = self.norm2(src)
|
||||
return src
|
||||
|
|
@ -0,0 +1,268 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
|
||||
import copy
|
||||
import math
|
||||
|
||||
import torch
|
||||
import torch.nn.functional as F
|
||||
from torch import Tensor, nn
|
||||
|
||||
|
||||
def _get_clones(module, N, layer_share=False):
|
||||
# import ipdb; ipdb.set_trace()
|
||||
if layer_share:
|
||||
return nn.ModuleList([module for i in range(N)])
|
||||
else:
|
||||
return nn.ModuleList([copy.deepcopy(module) for i in range(N)])
|
||||
|
||||
|
||||
def get_sine_pos_embed(
|
||||
pos_tensor: torch.Tensor,
|
||||
num_pos_feats: int = 128,
|
||||
temperature: int = 10000,
|
||||
exchange_xy: bool = True,
|
||||
):
|
||||
"""generate sine position embedding from a position tensor
|
||||
Args:
|
||||
pos_tensor (torch.Tensor): shape: [..., n].
|
||||
num_pos_feats (int): projected shape for each float in the tensor.
|
||||
temperature (int): temperature in the sine/cosine function.
|
||||
exchange_xy (bool, optional): exchange pos x and pos y. \
|
||||
For example, input tensor is [x,y], the results will be [pos(y), pos(x)]. Defaults to True.
|
||||
Returns:
|
||||
pos_embed (torch.Tensor): shape: [..., n*num_pos_feats].
|
||||
"""
|
||||
scale = 2 * math.pi
|
||||
dim_t = torch.arange(num_pos_feats, dtype=torch.float32, device=pos_tensor.device)
|
||||
dim_t = temperature ** (2 * torch.div(dim_t, 2, rounding_mode="floor") / num_pos_feats)
|
||||
|
||||
def sine_func(x: torch.Tensor):
|
||||
sin_x = x * scale / dim_t
|
||||
sin_x = torch.stack((sin_x[..., 0::2].sin(), sin_x[..., 1::2].cos()), dim=3).flatten(2)
|
||||
return sin_x
|
||||
|
||||
pos_res = [sine_func(x) for x in pos_tensor.split([1] * pos_tensor.shape[-1], dim=-1)]
|
||||
if exchange_xy:
|
||||
pos_res[0], pos_res[1] = pos_res[1], pos_res[0]
|
||||
pos_res = torch.cat(pos_res, dim=-1)
|
||||
return pos_res
|
||||
|
||||
|
||||
def gen_encoder_output_proposals(
|
||||
memory: Tensor, memory_padding_mask: Tensor, spatial_shapes: Tensor, learnedwh=None
|
||||
):
|
||||
"""
|
||||
Input:
|
||||
- memory: bs, \sum{hw}, d_model
|
||||
- memory_padding_mask: bs, \sum{hw}
|
||||
- spatial_shapes: nlevel, 2
|
||||
- learnedwh: 2
|
||||
Output:
|
||||
- output_memory: bs, \sum{hw}, d_model
|
||||
- output_proposals: bs, \sum{hw}, 4
|
||||
"""
|
||||
N_, S_, C_ = memory.shape
|
||||
proposals = []
|
||||
_cur = 0
|
||||
for lvl, (H_, W_) in enumerate(spatial_shapes):
|
||||
mask_flatten_ = memory_padding_mask[:, _cur : (_cur + H_ * W_)].view(N_, H_, W_, 1)
|
||||
valid_H = torch.sum(~mask_flatten_[:, :, 0, 0], 1)
|
||||
valid_W = torch.sum(~mask_flatten_[:, 0, :, 0], 1)
|
||||
|
||||
# import ipdb; ipdb.set_trace()
|
||||
|
||||
grid_y, grid_x = torch.meshgrid(
|
||||
torch.linspace(0, H_ - 1, H_, dtype=torch.float32, device=memory.device),
|
||||
torch.linspace(0, W_ - 1, W_, dtype=torch.float32, device=memory.device),
|
||||
)
|
||||
grid = torch.cat([grid_x.unsqueeze(-1), grid_y.unsqueeze(-1)], -1) # H_, W_, 2
|
||||
|
||||
scale = torch.cat([valid_W.unsqueeze(-1), valid_H.unsqueeze(-1)], 1).view(N_, 1, 1, 2)
|
||||
grid = (grid.unsqueeze(0).expand(N_, -1, -1, -1) + 0.5) / scale
|
||||
|
||||
if learnedwh is not None:
|
||||
# import ipdb; ipdb.set_trace()
|
||||
wh = torch.ones_like(grid) * learnedwh.sigmoid() * (2.0**lvl)
|
||||
else:
|
||||
wh = torch.ones_like(grid) * 0.05 * (2.0**lvl)
|
||||
|
||||
# scale = torch.cat([W_[None].unsqueeze(-1), H_[None].unsqueeze(-1)], 1).view(1, 1, 1, 2).repeat(N_, 1, 1, 1)
|
||||
# grid = (grid.unsqueeze(0).expand(N_, -1, -1, -1) + 0.5) / scale
|
||||
# wh = torch.ones_like(grid) / scale
|
||||
proposal = torch.cat((grid, wh), -1).view(N_, -1, 4)
|
||||
proposals.append(proposal)
|
||||
_cur += H_ * W_
|
||||
# import ipdb; ipdb.set_trace()
|
||||
output_proposals = torch.cat(proposals, 1)
|
||||
output_proposals_valid = ((output_proposals > 0.01) & (output_proposals < 0.99)).all(
|
||||
-1, keepdim=True
|
||||
)
|
||||
output_proposals = torch.log(output_proposals / (1 - output_proposals)) # unsigmoid
|
||||
output_proposals = output_proposals.masked_fill(memory_padding_mask.unsqueeze(-1), float("inf"))
|
||||
output_proposals = output_proposals.masked_fill(~output_proposals_valid, float("inf"))
|
||||
|
||||
output_memory = memory
|
||||
output_memory = output_memory.masked_fill(memory_padding_mask.unsqueeze(-1), float(0))
|
||||
output_memory = output_memory.masked_fill(~output_proposals_valid, float(0))
|
||||
|
||||
# output_memory = output_memory.masked_fill(memory_padding_mask.unsqueeze(-1), float('inf'))
|
||||
# output_memory = output_memory.masked_fill(~output_proposals_valid, float('inf'))
|
||||
|
||||
return output_memory, output_proposals
|
||||
|
||||
|
||||
class RandomBoxPerturber:
|
||||
def __init__(
|
||||
self, x_noise_scale=0.2, y_noise_scale=0.2, w_noise_scale=0.2, h_noise_scale=0.2
|
||||
) -> None:
|
||||
self.noise_scale = torch.Tensor(
|
||||
[x_noise_scale, y_noise_scale, w_noise_scale, h_noise_scale]
|
||||
)
|
||||
|
||||
def __call__(self, refanchors: Tensor) -> Tensor:
|
||||
nq, bs, query_dim = refanchors.shape
|
||||
device = refanchors.device
|
||||
|
||||
noise_raw = torch.rand_like(refanchors)
|
||||
noise_scale = self.noise_scale.to(device)[:query_dim]
|
||||
|
||||
new_refanchors = refanchors * (1 + (noise_raw - 0.5) * noise_scale)
|
||||
return new_refanchors.clamp_(0, 1)
|
||||
|
||||
|
||||
def sigmoid_focal_loss(
|
||||
inputs, targets, num_boxes, alpha: float = 0.25, gamma: float = 2, no_reduction=False
|
||||
):
|
||||
"""
|
||||
Loss used in RetinaNet for dense detection: https://arxiv.org/abs/1708.02002.
|
||||
Args:
|
||||
inputs: A float tensor of arbitrary shape.
|
||||
The predictions for each example.
|
||||
targets: A float tensor with the same shape as inputs. Stores the binary
|
||||
classification label for each element in inputs
|
||||
(0 for the negative class and 1 for the positive class).
|
||||
alpha: (optional) Weighting factor in range (0,1) to balance
|
||||
positive vs negative examples. Default = -1 (no weighting).
|
||||
gamma: Exponent of the modulating factor (1 - p_t) to
|
||||
balance easy vs hard examples.
|
||||
Returns:
|
||||
Loss tensor
|
||||
"""
|
||||
prob = inputs.sigmoid()
|
||||
ce_loss = F.binary_cross_entropy_with_logits(inputs, targets, reduction="none")
|
||||
p_t = prob * targets + (1 - prob) * (1 - targets)
|
||||
loss = ce_loss * ((1 - p_t) ** gamma)
|
||||
|
||||
if alpha >= 0:
|
||||
alpha_t = alpha * targets + (1 - alpha) * (1 - targets)
|
||||
loss = alpha_t * loss
|
||||
|
||||
if no_reduction:
|
||||
return loss
|
||||
|
||||
return loss.mean(1).sum() / num_boxes
|
||||
|
||||
|
||||
class MLP(nn.Module):
|
||||
"""Very simple multi-layer perceptron (also called FFN)"""
|
||||
|
||||
def __init__(self, input_dim, hidden_dim, output_dim, num_layers):
|
||||
super().__init__()
|
||||
self.num_layers = num_layers
|
||||
h = [hidden_dim] * (num_layers - 1)
|
||||
self.layers = nn.ModuleList(
|
||||
nn.Linear(n, k) for n, k in zip([input_dim] + h, h + [output_dim])
|
||||
)
|
||||
|
||||
def forward(self, x):
|
||||
for i, layer in enumerate(self.layers):
|
||||
x = F.relu(layer(x)) if i < self.num_layers - 1 else layer(x)
|
||||
return x
|
||||
|
||||
|
||||
def _get_activation_fn(activation, d_model=256, batch_dim=0):
|
||||
"""Return an activation function given a string"""
|
||||
if activation == "relu":
|
||||
return F.relu
|
||||
if activation == "gelu":
|
||||
return F.gelu
|
||||
if activation == "glu":
|
||||
return F.glu
|
||||
if activation == "prelu":
|
||||
return nn.PReLU()
|
||||
if activation == "selu":
|
||||
return F.selu
|
||||
|
||||
raise RuntimeError(f"activation should be relu/gelu, not {activation}.")
|
||||
|
||||
|
||||
def gen_sineembed_for_position(pos_tensor):
|
||||
# n_query, bs, _ = pos_tensor.size()
|
||||
# sineembed_tensor = torch.zeros(n_query, bs, 256)
|
||||
scale = 2 * math.pi
|
||||
dim_t = torch.arange(128, dtype=torch.float32, device=pos_tensor.device)
|
||||
dim_t = 10000 ** (2 * (torch.div(dim_t, 2, rounding_mode='floor')) / 128)
|
||||
x_embed = pos_tensor[:, :, 0] * scale
|
||||
y_embed = pos_tensor[:, :, 1] * scale
|
||||
pos_x = x_embed[:, :, None] / dim_t
|
||||
pos_y = y_embed[:, :, None] / dim_t
|
||||
pos_x = torch.stack((pos_x[:, :, 0::2].sin(), pos_x[:, :, 1::2].cos()), dim=3).flatten(2)
|
||||
pos_y = torch.stack((pos_y[:, :, 0::2].sin(), pos_y[:, :, 1::2].cos()), dim=3).flatten(2)
|
||||
if pos_tensor.size(-1) == 2:
|
||||
pos = torch.cat((pos_y, pos_x), dim=2)
|
||||
elif pos_tensor.size(-1) == 4:
|
||||
w_embed = pos_tensor[:, :, 2] * scale
|
||||
pos_w = w_embed[:, :, None] / dim_t
|
||||
pos_w = torch.stack((pos_w[:, :, 0::2].sin(), pos_w[:, :, 1::2].cos()), dim=3).flatten(2)
|
||||
|
||||
h_embed = pos_tensor[:, :, 3] * scale
|
||||
pos_h = h_embed[:, :, None] / dim_t
|
||||
pos_h = torch.stack((pos_h[:, :, 0::2].sin(), pos_h[:, :, 1::2].cos()), dim=3).flatten(2)
|
||||
|
||||
pos = torch.cat((pos_y, pos_x, pos_w, pos_h), dim=2)
|
||||
else:
|
||||
raise ValueError("Unknown pos_tensor shape(-1):{}".format(pos_tensor.size(-1)))
|
||||
return pos
|
||||
|
||||
|
||||
class ContrastiveEmbed(nn.Module):
|
||||
def __init__(self, max_text_len=256):
|
||||
"""
|
||||
Args:
|
||||
max_text_len: max length of text.
|
||||
"""
|
||||
super().__init__()
|
||||
self.max_text_len = max_text_len
|
||||
|
||||
def forward(self, x, text_dict):
|
||||
"""_summary_
|
||||
|
||||
Args:
|
||||
x (_type_): _description_
|
||||
text_dict (_type_): _description_
|
||||
{
|
||||
'encoded_text': encoded_text, # bs, 195, d_model
|
||||
'text_token_mask': text_token_mask, # bs, 195
|
||||
# True for used tokens. False for padding tokens
|
||||
}
|
||||
Returns:
|
||||
_type_: _description_
|
||||
"""
|
||||
assert isinstance(text_dict, dict)
|
||||
|
||||
y = text_dict["encoded_text"]
|
||||
text_token_mask = text_dict["text_token_mask"]
|
||||
|
||||
res = x @ y.transpose(-1, -2)
|
||||
res.masked_fill_(~text_token_mask[:, None, :], float("-inf"))
|
||||
|
||||
# padding to max_text_len
|
||||
new_res = torch.full((*res.shape[:-1], self.max_text_len), float("-inf"), device=res.device)
|
||||
new_res[..., : res.shape[-1]] = res
|
||||
|
||||
return new_res
|
||||
|
|
@ -0,0 +1,18 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
from .GroundingDINO import build_groundingdino
|
||||
|
||||
|
||||
def build_model(args, cfg):
|
||||
# we use register to maintain models from catdet6 on.
|
||||
from .registry import MODULE_BUILD_FUNCS
|
||||
|
||||
assert args.modelname in MODULE_BUILD_FUNCS._module_dict
|
||||
build_func = MODULE_BUILD_FUNCS.get(args.modelname)
|
||||
model = build_func(args, cfg)
|
||||
return model
|
||||
|
|
@ -0,0 +1,66 @@
|
|||
# ------------------------------------------------------------------------
|
||||
# Grounding DINO
|
||||
# url: https://github.com/IDEA-Research/GroundingDINO
|
||||
# Copyright (c) 2023 IDEA. All Rights Reserved.
|
||||
# Licensed under the Apache License, Version 2.0 [see LICENSE for details]
|
||||
# ------------------------------------------------------------------------
|
||||
# -*- coding: utf-8 -*-
|
||||
# @Author: Yihao Chen
|
||||
# @Date: 2021-08-16 16:03:17
|
||||
# @Last Modified by: Shilong Liu
|
||||
# @Last Modified time: 2022-01-23 15:26
|
||||
# modified from mmcv
|
||||
|
||||
import inspect
|
||||
from functools import partial
|
||||
|
||||
|
||||
class Registry(object):
|
||||
def __init__(self, name):
|
||||
self._name = name
|
||||
self._module_dict = dict()
|
||||
|
||||
def __repr__(self):
|
||||
format_str = self.__class__.__name__ + "(name={}, items={})".format(
|
||||
self._name, list(self._module_dict.keys())
|
||||
)
|
||||
return format_str
|
||||
|
||||
def __len__(self):
|
||||
return len(self._module_dict)
|
||||
|
||||
@property
|
||||
def name(self):
|
||||
return self._name
|
||||
|
||||
@property
|
||||
def module_dict(self):
|
||||
return self._module_dict
|
||||
|
||||
def get(self, key):
|
||||
return self._module_dict.get(key, None)
|
||||
|
||||
def registe_with_name(self, module_name=None, force=False):
|
||||
return partial(self.register, module_name=module_name, force=force)
|
||||
|
||||
def register(self, module_build_function, module_name=None, force=False):
|
||||
"""Register a module build function.
|
||||
Args:
|
||||
module (:obj:`nn.Module`): Module to be registered.
|
||||
"""
|
||||
if not inspect.isfunction(module_build_function):
|
||||
raise TypeError(
|
||||
"module_build_function must be a function, but got {}".format(
|
||||
type(module_build_function)
|
||||
)
|
||||
)
|
||||
if module_name is None:
|
||||
module_name = module_build_function.__name__
|
||||
if not force and module_name in self._module_dict:
|
||||
raise KeyError("{} is already registered in {}".format(module_name, self.name))
|
||||
self._module_dict[module_name] = module_build_function
|
||||
|
||||
return module_build_function
|
||||
|
||||
|
||||
MODULE_BUILD_FUNCS = Registry("model build functions")
|
||||
|
|
@ -0,0 +1,22 @@
|
|||
import argparse
|
||||
import os
|
||||
import sys
|
||||
|
||||
import numpy as np
|
||||
import torch
|
||||
from PIL import Image, ImageDraw, ImageFont
|
||||
|
||||
import groundingdino_new.datasets.transforms as T
|
||||
from groundingdino_new.models import build_model
|
||||
from groundingdino_new.util import box_ops
|
||||
from groundingdino_new.util.slconfig import SLConfig
|
||||
from groundingdino_new.util.utils import clean_state_dict, get_phrases_from_posmap
|
||||
|
||||
def load_model(model_config_path, model_checkpoint_path, cpu_only=False):
|
||||
args = SLConfig.fromfile(model_config_path)
|
||||
args.device = "cuda" if not cpu_only else "cpu"
|
||||
model = build_model(args)
|
||||
checkpoint = torch.load(model_checkpoint_path, map_location="cpu")
|
||||
load_res = model.load_state_dict(clean_state_dict(checkpoint["model"]), strict=False)
|
||||
print(load_res)
|
||||
return model
|
||||
|
|
@ -0,0 +1 @@
|
|||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
|
|
@ -0,0 +1,140 @@
|
|||
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved
|
||||
"""
|
||||
Utilities for bounding box manipulation and GIoU.
|
||||
"""
|
||||
import torch
|
||||
from torchvision.ops.boxes import box_area
|
||||
|
||||
|
||||
def box_cxcywh_to_xyxy(x):
|
||||
x_c, y_c, w, h = x.unbind(-1)
|
||||
b = [(x_c - 0.5 * w), (y_c - 0.5 * h), (x_c + 0.5 * w), (y_c + 0.5 * h)]
|
||||
return torch.stack(b, dim=-1)
|
||||
|
||||
|
||||
def box_xyxy_to_cxcywh(x):
|
||||
x0, y0, x1, y1 = x.unbind(-1)
|
||||
b = [(x0 + x1) / 2, (y0 + y1) / 2, (x1 - x0), (y1 - y0)]
|
||||
return torch.stack(b, dim=-1)
|
||||
|
||||
|
||||
# modified from torchvision to also return the union
|
||||
def box_iou(boxes1, boxes2):
|
||||
area1 = box_area(boxes1)
|
||||
area2 = box_area(boxes2)
|
||||
|
||||
# import ipdb; ipdb.set_trace()
|
||||
lt = torch.max(boxes1[:, None, :2], boxes2[:, :2]) # [N,M,2]
|
||||
rb = torch.min(boxes1[:, None, 2:], boxes2[:, 2:]) # [N,M,2]
|
||||
|
||||
wh = (rb - lt).clamp(min=0) # [N,M,2]
|
||||
inter = wh[:, :, 0] * wh[:, :, 1] # [N,M]
|
||||
|
||||
union = area1[:, None] + area2 - inter
|
||||
|
||||
iou = inter / (union + 1e-6)
|
||||
return iou, union
|
||||
|
||||
|
||||
def generalized_box_iou(boxes1, boxes2):
|
||||
"""
|
||||
Generalized IoU from https://giou.stanford.edu/
|
||||
|
||||
The boxes should be in [x0, y0, x1, y1] format
|
||||
|
||||
Returns a [N, M] pairwise matrix, where N = len(boxes1)
|
||||
and M = len(boxes2)
|
||||
"""
|
||||
# degenerate boxes gives inf / nan results
|
||||
# so do an early check
|
||||
assert (boxes1[:, 2:] >= boxes1[:, :2]).all()
|
||||
assert (boxes2[:, 2:] >= boxes2[:, :2]).all()
|
||||
# except:
|
||||
# import ipdb; ipdb.set_trace()
|
||||
iou, union = box_iou(boxes1, boxes2)
|
||||
|
||||
lt = torch.min(boxes1[:, None, :2], boxes2[:, :2])
|
||||
rb = torch.max(boxes1[:, None, 2:], boxes2[:, 2:])
|
||||
|
||||
wh = (rb - lt).clamp(min=0) # [N,M,2]
|
||||
area = wh[:, :, 0] * wh[:, :, 1]
|
||||
|
||||
return iou - (area - union) / (area + 1e-6)
|
||||
|
||||
|
||||
# modified from torchvision to also return the union
|
||||
def box_iou_pairwise(boxes1, boxes2):
|
||||
area1 = box_area(boxes1)
|
||||
area2 = box_area(boxes2)
|
||||
|
||||
lt = torch.max(boxes1[:, :2], boxes2[:, :2]) # [N,2]
|
||||
rb = torch.min(boxes1[:, 2:], boxes2[:, 2:]) # [N,2]
|
||||
|
||||
wh = (rb - lt).clamp(min=0) # [N,2]
|
||||
inter = wh[:, 0] * wh[:, 1] # [N]
|
||||
|
||||
union = area1 + area2 - inter
|
||||
|
||||
iou = inter / union
|
||||
return iou, union
|
||||
|
||||
|
||||
def generalized_box_iou_pairwise(boxes1, boxes2):
|
||||
"""
|
||||
Generalized IoU from https://giou.stanford.edu/
|
||||
|
||||
Input:
|
||||
- boxes1, boxes2: N,4
|
||||
Output:
|
||||
- giou: N, 4
|
||||
"""
|
||||
# degenerate boxes gives inf / nan results
|
||||
# so do an early check
|
||||
assert (boxes1[:, 2:] >= boxes1[:, :2]).all()
|
||||
assert (boxes2[:, 2:] >= boxes2[:, :2]).all()
|
||||
assert boxes1.shape == boxes2.shape
|
||||
iou, union = box_iou_pairwise(boxes1, boxes2) # N, 4
|
||||
|
||||
lt = torch.min(boxes1[:, :2], boxes2[:, :2])
|
||||
rb = torch.max(boxes1[:, 2:], boxes2[:, 2:])
|
||||
|
||||
wh = (rb - lt).clamp(min=0) # [N,2]
|
||||
area = wh[:, 0] * wh[:, 1]
|
||||
|
||||
return iou - (area - union) / area
|
||||
|
||||
|
||||
def masks_to_boxes(masks):
|
||||
"""Compute the bounding boxes around the provided masks
|
||||
|
||||
The masks should be in format [N, H, W] where N is the number of masks, (H, W) are the spatial dimensions.
|
||||
|
||||
Returns a [N, 4] tensors, with the boxes in xyxy format
|
||||
"""
|
||||
if masks.numel() == 0:
|
||||
return torch.zeros((0, 4), device=masks.device)
|
||||
|
||||
h, w = masks.shape[-2:]
|
||||
|
||||
y = torch.arange(0, h, dtype=torch.float)
|
||||
x = torch.arange(0, w, dtype=torch.float)
|
||||
y, x = torch.meshgrid(y, x)
|
||||
|
||||
x_mask = masks * x.unsqueeze(0)
|
||||
x_max = x_mask.flatten(1).max(-1)[0]
|
||||
x_min = x_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0]
|
||||
|
||||
y_mask = masks * y.unsqueeze(0)
|
||||
y_max = y_mask.flatten(1).max(-1)[0]
|
||||
y_min = y_mask.masked_fill(~(masks.bool()), 1e8).flatten(1).min(-1)[0]
|
||||
|
||||
return torch.stack([x_min, y_min, x_max, y_max], 1)
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
x = torch.rand(5, 4)
|
||||
y = torch.rand(3, 4)
|
||||
iou, union = box_iou(x, y)
|
||||
import ipdb
|
||||
|
||||
ipdb.set_trace()
|
||||
|
|
@ -0,0 +1,26 @@
|
|||
from transformers import AutoTokenizer, BertModel, BertTokenizer, RobertaModel, RobertaTokenizerFast
|
||||
|
||||
|
||||
def get_tokenlizer(text_encoder_type):
|
||||
if not isinstance(text_encoder_type, str):
|
||||
# print("text_encoder_type is not a str")
|
||||
if hasattr(text_encoder_type, "text_encoder_type"):
|
||||
text_encoder_type = text_encoder_type.text_encoder_type
|
||||
elif text_encoder_type.get("text_encoder_type", False):
|
||||
text_encoder_type = text_encoder_type.get("text_encoder_type")
|
||||
else:
|
||||
raise ValueError(
|
||||
"Unknown type of text_encoder_type: {}".format(type(text_encoder_type))
|
||||
)
|
||||
print("final text_encoder_type: {}".format(text_encoder_type))
|
||||
|
||||
tokenizer = AutoTokenizer.from_pretrained(text_encoder_type)
|
||||
return tokenizer
|
||||
|
||||
|
||||
def get_pretrained_language_model(text_encoder_type):
|
||||
if "bert-base-uncased" in text_encoder_type:
|
||||
return BertModel.from_pretrained(text_encoder_type)
|
||||
if text_encoder_type == "roberta-base":
|
||||
return RobertaModel.from_pretrained(text_encoder_type)
|
||||
raise ValueError("Unknown text_encoder_type {}".format(text_encoder_type))
|
||||
|
|
@ -0,0 +1,242 @@
|
|||
from typing import Tuple, List
|
||||
|
||||
import cv2
|
||||
import numpy as np
|
||||
import supervision as sv
|
||||
import torch
|
||||
from PIL import Image
|
||||
from torchvision.ops import box_convert
|
||||
|
||||
import groundingdino_new.datasets.transforms as T
|
||||
from groundingdino_new.models import build_model
|
||||
from groundingdino_new.util.misc import clean_state_dict
|
||||
from groundingdino_new.util.slconfig import SLConfig
|
||||
from groundingdino_new.util.utils import get_phrases_from_posmap
|
||||
|
||||
# ----------------------------------------------------------------------------------------------------------------------
|
||||
# OLD API
|
||||
# ----------------------------------------------------------------------------------------------------------------------
|
||||
|
||||
|
||||
def preprocess_caption(caption: str) -> str:
|
||||
result = caption.lower().strip()
|
||||
if result.endswith("."):
|
||||
return result
|
||||
return result + "."
|
||||
|
||||
|
||||
def load_model(model_config_path: str, model_checkpoint_path: str, device: str = "cuda"):
|
||||
args = SLConfig.fromfile(model_config_path)
|
||||
args.device = device
|
||||
model = build_model(args)
|
||||
checkpoint = torch.load(model_checkpoint_path, map_location="cpu")
|
||||
model.load_state_dict(clean_state_dict(checkpoint["model"]), strict=False)
|
||||
model.eval()
|
||||
return model
|
||||
|
||||
|
||||
def load_image(image_path: str) -> Tuple[np.array, torch.Tensor]:
|
||||
transform = T.Compose(
|
||||
[
|
||||
T.RandomResize([800], max_size=1333),
|
||||
T.ToTensor(),
|
||||
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
|
||||
]
|
||||
)
|
||||
image_source = Image.open(image_path).convert("RGB")
|
||||
image = np.asarray(image_source)
|
||||
image_transformed, _ = transform(image_source, None)
|
||||
return image, image_transformed
|
||||
|
||||
|
||||
def predict(
|
||||
model,
|
||||
image: torch.Tensor,
|
||||
caption: str,
|
||||
box_threshold: float,
|
||||
text_threshold: float,
|
||||
device: str = "cuda"
|
||||
) -> Tuple[torch.Tensor, torch.Tensor, List[str]]:
|
||||
caption = preprocess_caption(caption=caption)
|
||||
|
||||
model = model.to(device)
|
||||
image = image.to(device)
|
||||
|
||||
with torch.no_grad():
|
||||
outputs = model(image[None], captions=[caption])
|
||||
|
||||
prediction_logits = outputs["pred_logits"].cpu().sigmoid()[0] # prediction_logits.shape = (nq, 256)
|
||||
prediction_boxes = outputs["pred_boxes"].cpu()[0] # prediction_boxes.shape = (nq, 4)
|
||||
|
||||
mask = prediction_logits.max(dim=1)[0] > box_threshold
|
||||
logits = prediction_logits[mask] # logits.shape = (n, 256)
|
||||
boxes = prediction_boxes[mask] # boxes.shape = (n, 4)
|
||||
|
||||
tokenizer = model.tokenizer
|
||||
tokenized = tokenizer(caption)
|
||||
|
||||
phrases = [
|
||||
get_phrases_from_posmap(logit > text_threshold, tokenized, tokenizer).replace('.', '')
|
||||
for logit
|
||||
in logits
|
||||
]
|
||||
|
||||
return boxes, logits.max(dim=1)[0], phrases
|
||||
|
||||
|
||||
def annotate(image_source: np.ndarray, boxes: torch.Tensor, logits: torch.Tensor, phrases: List[str]) -> np.ndarray:
|
||||
h, w, _ = image_source.shape
|
||||
boxes = boxes * torch.Tensor([w, h, w, h])
|
||||
xyxy = box_convert(boxes=boxes, in_fmt="cxcywh", out_fmt="xyxy").numpy()
|
||||
detections = sv.Detections(xyxy=xyxy)
|
||||
|
||||
labels = [
|
||||
f"{phrase} {logit:.2f}"
|
||||
for phrase, logit
|
||||
in zip(phrases, logits)
|
||||
]
|
||||
|
||||
box_annotator = sv.BoxAnnotator()
|
||||
annotated_frame = cv2.cvtColor(image_source, cv2.COLOR_RGB2BGR)
|
||||
annotated_frame = box_annotator.annotate(scene=annotated_frame, detections=detections, labels=labels)
|
||||
return annotated_frame
|
||||
|
||||
|
||||
# ----------------------------------------------------------------------------------------------------------------------
|
||||
# NEW API
|
||||
# ----------------------------------------------------------------------------------------------------------------------
|
||||
|
||||
|
||||
class Model:
|
||||
|
||||
def __init__(
|
||||
self,
|
||||
model_config_path: str,
|
||||
model_checkpoint_path: str,
|
||||
device: str = "cuda"
|
||||
):
|
||||
self.model = load_model(
|
||||
model_config_path=model_config_path,
|
||||
model_checkpoint_path=model_checkpoint_path,
|
||||
device=device
|
||||
).to(device)
|
||||
self.device = device
|
||||
|
||||
def predict_with_caption(
|
||||
self,
|
||||
image: np.ndarray,
|
||||
caption: str,
|
||||
box_threshold: float = 0.35,
|
||||
text_threshold: float = 0.25
|
||||
) -> Tuple[sv.Detections, List[str]]:
|
||||
"""
|
||||
import cv2
|
||||
|
||||
image = cv2.imread(IMAGE_PATH)
|
||||
|
||||
model = Model(model_config_path=CONFIG_PATH, model_checkpoint_path=WEIGHTS_PATH)
|
||||
detections, labels = model.predict_with_caption(
|
||||
image=image,
|
||||
caption=caption,
|
||||
box_threshold=BOX_THRESHOLD,
|
||||
text_threshold=TEXT_THRESHOLD
|
||||
)
|
||||
|
||||
import supervision as sv
|
||||
|
||||
box_annotator = sv.BoxAnnotator()
|
||||
annotated_image = box_annotator.annotate(scene=image, detections=detections, labels=labels)
|
||||
"""
|
||||
processed_image = Model.preprocess_image(image_bgr=image).to(self.device)
|
||||
boxes, logits, phrases = predict(
|
||||
model=self.model,
|
||||
image=processed_image,
|
||||
caption=caption,
|
||||
box_threshold=box_threshold,
|
||||
text_threshold=text_threshold)
|
||||
source_h, source_w, _ = image.shape
|
||||
detections = Model.post_process_result(
|
||||
source_h=source_h,
|
||||
source_w=source_w,
|
||||
boxes=boxes,
|
||||
logits=logits)
|
||||
return detections, phrases
|
||||
|
||||
def predict_with_classes(
|
||||
self,
|
||||
image: np.ndarray,
|
||||
classes: List[str],
|
||||
box_threshold: float,
|
||||
text_threshold: float
|
||||
) -> sv.Detections:
|
||||
"""
|
||||
import cv2
|
||||
|
||||
image = cv2.imread(IMAGE_PATH)
|
||||
|
||||
model = Model(model_config_path=CONFIG_PATH, model_checkpoint_path=WEIGHTS_PATH)
|
||||
detections = model.predict_with_classes(
|
||||
image=image,
|
||||
classes=CLASSES,
|
||||
box_threshold=BOX_THRESHOLD,
|
||||
text_threshold=TEXT_THRESHOLD
|
||||
)
|
||||
|
||||
|
||||
import supervision as sv
|
||||
|
||||
box_annotator = sv.BoxAnnotator()
|
||||
annotated_image = box_annotator.annotate(scene=image, detections=detections)
|
||||
"""
|
||||
caption = ", ".join(classes)
|
||||
processed_image = Model.preprocess_image(image_bgr=image).to(self.device)
|
||||
boxes, logits, phrases = predict(
|
||||
model=self.model,
|
||||
image=processed_image,
|
||||
caption=caption,
|
||||
box_threshold=box_threshold,
|
||||
text_threshold=text_threshold)
|
||||
source_h, source_w, _ = image.shape
|
||||
detections = Model.post_process_result(
|
||||
source_h=source_h,
|
||||
source_w=source_w,
|
||||
boxes=boxes,
|
||||
logits=logits)
|
||||
class_id = Model.phrases2classes(phrases=phrases, classes=classes)
|
||||
detections.class_id = class_id
|
||||
return detections
|
||||
|
||||
@staticmethod
|
||||
def preprocess_image(image_bgr: np.ndarray) -> torch.Tensor:
|
||||
transform = T.Compose(
|
||||
[
|
||||
T.RandomResize([800], max_size=1333),
|
||||
T.ToTensor(),
|
||||
T.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225]),
|
||||
]
|
||||
)
|
||||
image_pillow = Image.fromarray(cv2.cvtColor(image_bgr, cv2.COLOR_BGR2RGB))
|
||||
image_transformed, _ = transform(image_pillow, None)
|
||||
return image_transformed
|
||||
|
||||
@staticmethod
|
||||
def post_process_result(
|
||||
source_h: int,
|
||||
source_w: int,
|
||||
boxes: torch.Tensor,
|
||||
logits: torch.Tensor
|
||||
) -> sv.Detections:
|
||||
boxes = boxes * torch.Tensor([source_w, source_h, source_w, source_h])
|
||||
xyxy = box_convert(boxes=boxes, in_fmt="cxcywh", out_fmt="xyxy").numpy()
|
||||
confidence = logits.numpy()
|
||||
return sv.Detections(xyxy=xyxy, confidence=confidence)
|
||||
|
||||
@staticmethod
|
||||
def phrases2classes(phrases: List[str], classes: List[str]) -> np.ndarray:
|
||||
class_ids = []
|
||||
for phrase in phrases:
|
||||
try:
|
||||
class_ids.append(classes.index(phrase))
|
||||
except ValueError:
|
||||
class_ids.append(None)
|
||||
return np.array(class_ids)
|
||||
Some files were not shown because too many files have changed in this diff Show More
Loading…
Reference in New Issue