mirrors
/
reid-strong-baseline
mirror of https://github.com/michuanhaohao/reid-strong-baseline
1
0
Fork 0
Code Issues Projects Releases Wiki Activity

update readme

pull/5/head
michuanhaohao 2019-03-17 21:37:15 +08:00
parent 1a1f4eb5e0
commit 5252b355f2
2 changed files with 23 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
README.md
View File

@ -15,6 +15,26 @@ Bag of tricks
- BNNeck - BNNeck
- Center loss - Center loss
## Pipeline
<div align=center>
<img src='imgs/pipeline.jpg' width='800'>
</div>
## Results (rank1/mAP)
| Model | Market1501 | DukeMTMC-reID |
| --- | -- | -- |
| Standard baseline | 87.7 (74.0) | 79.7 (63.8) |
| +Warmup | 88.7 (75.2) | 80.6(65.1) |
| +Random erasing augmentation | 91.3 (79.3) | 81.5 (68.3) |
| +Label smoothing | 91.4 (80.3) | 82.4 (69.3) |
| +Last stride=1 | 92.0 (81.7) | 82.6 (70.6) |
| +BNNeck | 94.1 (85.7) | 86.2 (75.9) |
| +Center loss | 94.5 (85.9) | 86.4 (76.4) |
| +Reranking | 95.4 (94.2) | 90.3 (89.1) |
[model(Market1501)]()
[model(DukeMTMC-reID)]()
## Get Started ## Get Started
The designed architecture follows this guide [PyTorch-Project-Template](https://github.com/L1aoXingyu/PyTorch-Project-Template), you can check each folder's purpose by yourself. The designed architecture follows this guide [PyTorch-Project-Template](https://github.com/L1aoXingyu/PyTorch-Project-Template), you can check each folder's purpose by yourself.
@ -23,7 +43,7 @@ The designed architecture follows this guide [PyTorch-Project-Template](https://
2. Run `git clone... ` 2. Run `git clone... `
3. Install dependencies: 3. Install dependencies:
- [pytorch 1.0](https://pytorch.org/) - [pytorch>=0.4](https://pytorch.org/)
- torchvision - torchvision
- [ignite](https://github.com/pytorch/ignite) - [ignite](https://github.com/pytorch/ignite)
- [yacs](https://github.com/rbgirshick/yacs) - [yacs](https://github.com/rbgirshick/yacs)
@ -89,6 +109,8 @@ python3 tools/train.py --config_file='configs/softmax_triplet_with_center.yml' M
## Test ## Test
You can test your model's performance directly by running these commands in `.sh ` files. You can also change the configuration to determine which feature of BNNeck and whether the feature is normalized (equivalent to use Cosine distance or Euclidean distance) for testing. You can test your model's performance directly by running these commands in `.sh ` files. You can also change the configuration to determine which feature of BNNeck and whether the feature is normalized (equivalent to use Cosine distance or Euclidean distance) for testing.
Please replace the data path of the model.
1. Test with Euclidean distance using feature before BN without re-ranking,. 1. Test with Euclidean distance using feature before BN without re-ranking,.
```bash ```bash
@ -105,8 +127,5 @@ python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MO
python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('your device id')" DATASETS.NAMES "('dukemtmc')" TEST.NECK_FEAT "('after')" TEST.FEAT_NORM "('yes')" TEST.RE_RANKING "('yes')" TEST.WEIGHT "('your path to trained checkpoints')" python3 tools/test.py --config_file='configs/softmax_triplet_with_center.yml' MODEL.DEVICE_ID "('your device id')" DATASETS.NAMES "('dukemtmc')" TEST.NECK_FEAT "('after')" TEST.FEAT_NORM "('yes')" TEST.RE_RANKING "('yes')" TEST.WEIGHT "('your path to trained checkpoints')"
``` ```
## Results
**Network architecture**

BIN
imgs/pipeline.jpg 100644
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 216 KiB