mirrors
/
mmselfsup
mirror of https://github.com/open-mmlab/mmselfsup.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
mmselfsup/docs_zh-CN/model_zoo.md

86 lines
12 KiB
Markdown
Raw Blame History

# Model Zoo
All models and benchmarks results are recorded below.
## Pre-trained models
| Algorithm | Config | Download | Remark |
| ----------------------------------------------------------------- | ------------------------------------------------------------------------------------------ | --------- | --------------------- |
| [BYOL](../configs/selfsup/byol/README.md) | [config](../configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k.py) | [model]() | 8xb32 / Accumulate 16 |
| [DeepCLuster](../configs/selfsup/deepcluster/README.md) | [config](../configs/selfsup/deepcluster/deepcluster_resnet50_8xb64-steplr-200e_in1k.py) | [model]() | |
| [DenseCL](../configs/selfsup/densecl/README.md) | [config](../configs/selfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | |
| [MoCo v2](../configs/selfsup/moco/README.md) | [config](../configs/selfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | |
| [NPID](../configs/selfsup/npid/README.md) | [config](../configs/selfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | |
| | [config](../configs/selfsup/npid/npid-ensure-neg_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | Ensure Negative |
| [Online DeepCluster](../configs/selfsup/odc/README.md) | [config](../configs/selfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k.py) | [model]() | |
| [Relative Location](../configs/selfsup/relative_loc/README.md) | [config](../configs/selfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k.py) | [model]() | |
| [Rotation Prediction](../configs/selfsup/rotation_pred/README.md) | [config](../configs/selfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k.py) | [model]() | |
| [SimCLR](../configs/selfsup/simclr/README.md) | [config](../configs/selfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | |
| [SimSiam](../configs/selfsup/simsiam/README.md) | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py) | [model]() | Epoch 100 |
| | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | Epoch 200 |
| [SwAV](../configs/selfsup/swav/README.md) | [config](../configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py) | [model]() | |
## Benchmarks
In following tables, we only displayed ImageNet Linear Evaluation, COCO17 Object Detection and PASCAL VOC12 Aug Segmentation, you can click the model name above to get the comprehensive benchmark results.
### ImageNet Linear Evaluation
If not specified, we use linear evaluation setting from [MoCo](http://openaccess.thecvf.com/content_CVPR_2020/papers/He_Momentum_Contrast_for_Unsupervised_Visual_Representation_Learning_CVPR_2020_paper.pdf). Or the settings is mentioned in Remark.
| Algorithm | Config | Download | Remark | Top-1 (%) |
| ------------------- | ------------------------------------------------------------------------------------------ | --------- | --------------------------------- | --------- |
| BYOL | [config](../configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k.py) | [model]() | 8xb32 / Accumulate 16 | 67.68 | |
| DeepCLuster | [config](../configs/selfsup/deepcluster/deepcluster_resnet50_8xb64-steplr-200e_in1k.py) | [model]() | | |
| DenseCL | [config](../configs/selfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| MoCo v2 | [config](../configs/selfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | 67.56 |
| NPID | [config](../configs/selfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | | |
| | [config](../configs/selfsup/npid/npid-ensure-neg_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | Ensure Negative | |
| Online DeepCluster | [config](../configs/selfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k.py) | [model]() | | |
| Relative Location | [config](../configs/selfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k.py) | [model]() | | 39.65 |
| Rotation Prediction | [config](../configs/selfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k.py) | [model]() | | 44.35 |
| SimCLR | [config](../configs/selfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| SimSiam | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py) | [model]() | Epoch 100 / SimSiam paper setting | |
| | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | Epoch 200 / SimSiam paper setting | |
| SwAV | [config](../configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py) | [model]() | SwAV paper setting | |
### COCO17 Object Detection
In COCO17 Object detection task, we choose the evluation protocol from [MoCo](http://openaccess.thecvf.com/content_CVPR_2020/papers/He_Momentum_Contrast_for_Unsupervised_Visual_Representation_Learning_CVPR_2020_paper.pdf), with Mask-RCNN architecture, the results below are trained with the same [config](../configs/benchmarks/mmdetection/coco/mask_rcnn_r50_fpn_mstrain_1x.py).
| Algorithm | Config | Download | Remark | AP (Box) | AP (Mask) |
| ------------------- | ------------------------------------------------------------------------------------------ | --------- | --------------------- | -------- | --------- |
| BYOL | [config](../configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k.py) | [model]() | 8xb32 / Accumulate 16 | |
| DeepCLuster | [config](../configs/selfsup/deepcluster/deepcluster_resnet50_8xb64-steplr-200e_in1k.py) | [model]() | | |
| DenseCL | [config](../configs/selfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| MoCo v2 | [config](../configs/selfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| NPID | [config](../configs/selfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | | |
| | [config](../configs/selfsup/npid/npid-ensure-neg_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | Ensure Negative | |
| Online DeepCluster | [config](../configs/selfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k.py) | [model]() | | |
| Relative Location | [config](../configs/selfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k.py) | [model]() | | |
| Rotation Prediction | [config](../configs/selfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k.py) | [model]() | | |
| SimCLR | [config](../configs/selfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| SimSiam | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py) | [model]() | Epoch 100 | 36.9 | 33.2 |
| | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | Epoch 200 | |
| SwAV | [config](../configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py) | [model]() | | |
### Pascal VOC12 Aug Segmentation
In Pascal VOC12 Aug Segmentation task, we choose the evluation protocol from [MoCo](http://openaccess.thecvf.com/content_CVPR_2020/papers/He_Momentum_Contrast_for_Unsupervised_Visual_Representation_Learning_CVPR_2020_paper.pdf), with FCN architecture, the results below are trained with the same [config](../configs/benchmarks/mmsegmentation/voc12aug/fcn_d6_r50-d16_512x512_30k_moco.py).
| Algorithm | Config | Download | Remark | mIOU |
| ------------------- | ------------------------------------------------------------------------------------------ | --------- | --------------------- | ---- |
| BYOL | [config](../configs/selfsup/byol/byol_resnet50_8xb32-accum16-coslr-200e_in1k.py) | [model]() | 8xb32 / Accumulate 16 | |
| DeepCLuster | [config](../configs/selfsup/deepcluster/deepcluster_resnet50_8xb64-steplr-200e_in1k.py) | [model]() | | |
| DenseCL | [config](../configs/selfsup/densecl/densecl_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| MoCo v2 | [config](../configs/selfsup/moco/mocov2_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| NPID | [config](../configs/selfsup/npid/npid_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | | |
| | [config](../configs/selfsup/npid/npid-ensure-neg_resnet50_8xb32-steplr-200e_in1k.py) | [model]() | Ensure Negative | |
| Online DeepCluster | [config](../configs/selfsup/odc/odc_resnet50_8xb64-steplr-440e_in1k.py) | [model]() | | |
| Relative Location | [config](../configs/selfsup/relative_loc/relative-loc_resnet50_8xb64-steplr-70e_in1k.py) | [model]() | | |
| Rotation Prediction | [config](../configs/selfsup/rotation_pred/rotation-pred_resnet50_8xb16-steplr-70e_in1k.py) | [model]() | | |
| SimCLR | [config](../configs/selfsup/simclr/simclr_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | | |
| SimSiam | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-100e_in1k.py) | [model]() | Epoch 100 | |
| | [config](../configs/selfsup/simsiam/simsiam_resnet50_8xb32-coslr-200e_in1k.py) | [model]() | Epoch 200 | |
| SwAV | [config](../configs/selfsup/swav/swav_resnet50_8xb32-mcrop-2-6-coslr-200e_in1k-224-96.py) | [model]() | | |
Reference in New Issue View Git Blame Copy Permalink