ECN

Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification CVPR 2019

Preparation

Requirements: Python=3.6 and Pytorch>=1.0.0

  1. Install Pytorch

  2. Download dataset

  • Market-1501 [BaiduYun] [GoogleDriver] CamStyle [GoogleDriver] [BaiduYun] (password: 6bu4)

  • DukeMTMC-reID [BaiduYun] (password: bhbh) [GoogleDriver] CamStyle [GoogleDriver] [BaiduYun] (password: 6bu4)

  • MSMT17 + CamStyle [BaiduYun] (password: 6bu4) [GoogleDriver] We reformulate the structure of MSMT17 the same as Market-1501.

  • Unzip each dataset and corresponding CamStyle under 'ECN/data/'

    Ensure the File structure is as follow:

    ECN/data    
    │
    └───market OR duke OR msmt17
       │   
       └───bounding_box_train
       │   
       └───bounding_box_test
       │   
       └───bounding_box_train_camstyle
       | 
       └───query
    

Training and test domain adaptation model for person re-ID

# For Duke to Market-1501
python main.py -s duke -t market --logs-dir logs/duke2market-ECN

# For Market-1501 to Duke
python main.py -s market -t duke --logs-dir logs/market2duke-ECN

# For Market-1501 to MSMT17
python main.py -s market -t msmt17 --logs-dir logs/market2msmt17-ECN --re 0

# For Duke to MSMT17
python main.py -s duke -t msmt17 --logs-dir logs/duke2msmt17-ECN --re 0

Results

market_duke

msmt17

References

  • [1] Our code is conducted based on open-reid

  • [2] Camera Style Adaptation for Person Re-identification. CVPR 2018.

  • [3] Generalizing A Person Retrieval Model Hetero- and Homogeneously. ECCV 2018.

Citation

If you find this code useful in your research, please consider citing:

@inproceedings{zhong2019invariance,
  title={Invariance Matters: Exemplar Memory for Domain Adaptive Person Re-identification},
  author={Zhong, Zhun and Zheng, Liang and Luo, Zhiming and Li, Shaozi and Yang, Yi},
  booktitle={Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
  year={2019},
}

GitHub