Prerequisites

• Python 3.6
• GPU Memory >= 11G
• 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

├── feature_disentangle
	　　　　　　   ├── 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

Finally, conduct training, testing and evaluating with one command

python run.py

This code is related to our paper A Feature Disentangling Approach for Person Re-Identification via Self-Supervised Data Augmentation.

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

@article{chen2021feature,
  title={A feature disentangling approach for person re-identification via self-supervised data augmentation},
  author={Chen, Feng and Wang, Nian and Tang, Jun and Zhu, Fan},
  journal={Applied Soft Computing},
  volume={100},
  pages={106939},
  year={2021},
  publisher={Elsevier}
}