Prerequisites

• Python 3.6
• GPU Memory >= 6G
• Numpy
• Pytorch 0.4+

Preparation 1: create folder for dataset.

first, download Market-1501 dataset from the links below:

google drive: https://drive.google.com/file/d/0B8-rUzbwVRk0c054eEozWG9COHM/view?usp=sharing

baidu disk: https://pan.baidu.com/s/1ntIi2Op

second,

mkdir data
unzip Market-1501-v15.09.15.zip
ln -s Market-1501-v15.09.15 market

then, get the directory structure

├── data_aug_reid
        ├── data
                ├── market
                ├── Market-1501-v15.09.15

Preparation 2: Put the images with the same id in one folder. You may use

python prepare.py
python changeIndex.py

Then, conduct ID data augmentation

python self_supervise_augment.py

Finally, conduct training, testing and evaluating with one command

python run.py

This code is related to our paper Self-supervised Data Augmentation for Person Re-Identification.

If you use this code, please cite our paper as:

@article{chen2020self,
  title={Self-supervised data augmentation for person re-identification},
  author={Chen, Feng and Wang, Nian and Tang, Jun and Liang, Dong and Feng, Hao},
  journal={Neurocomputing},
  volume={415},
  pages={48--59},
  year={2020},
  publisher={Elsevier}
}

reference link:

https://github.com/layumi/Person_reID_baseline_pytorch