Domain Generalization Person ReID Baseline

This repository is the code for the Baseline in Style Variable and Irrelevant Learning for Generalizable Person Re-identification. The overall code will be released later.

Requirements

Installation

git clone ...
cd dgreid/reid/evaluation_metrics/rank_cylib && make all

Prepare Datasets

Download the person re-ID datasets Market-1501, DukeMTMC-ReID, MSMT17, and cuhk03. Then unzip them under the root directory like

/data/datasets/
├── dukemtmc
│   └── DukeMTMC-reID
├── market1501
│   └── Market-1501-v15.09.15
├── msmt17
│   └── MSMT17_V1
├── cuhk03
    └── cuhk03_release

Training

By default we utilize 4 GTX-2080TI GPUs for training. Note that

• The multi-source domains are trained parallel with DP.
• More details of configs in reid/config/default_parser.py

Quickly Start

To train the baseline methods, run commands like:

# Base Baseline (w/o meta learning)
CUDA_VISIBLE_DEVICES=0,1,2,3 sh scripts/base_baseline.sh

# Strong Baseline (w/ Meta learning)
CUDA_VISIBLE_DEVICES=0,1,2,3 sh scripts/meta_baseline.sh

Citation

If you find our work is useful for your research, please kindly cite our paper

@misc{chen2022style,
      title={Style Variable and Irrelevant Learning for Generalizable Person Re-identification}, 
      author={Haobo Chen and Chuyang Zhao and Kai Tu and Junru Chen and Yadong Li and Boxun Li},
      year={2022},
      eprint={2209.05235},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

If you have any questions, please leave an issue or contact us: [email protected] or [email protected]

Acknowledgement

Our code is based on MMT and IDM.