• Learning to Disentangle Scenes for Person Re-identification
• • Getting Started
  • • Requirements
    • Download pre-trained model
  • Training and Testing
  • Results in the Paper
  • Acknowledgement
  • Citation
  • License

Learning to Disentangle Scenes for Person Re-identification

Implementation of the proposed LDS. For the preprint version, please refer to [Arxiv].

Getting Started

Requirements

Here is a brief instruction for installing the experimental environment.

# install virtual envs
$ conda create -n LDS python=3.7 -y
$ conda activate LDS
# install pytorch 1.6.0 and other dependencies

Download pre-trained model

The pre-trained vit model can be downloaded in this link(code: 7qup) and should be put in the /home/[USER]/.cache/torch/checkpoints/ directory.

Training and Testing

$ python projects\LDS\train_net.py --config-file projects/LDS/configs/Market1501/LDS_3Branch_mutual.yml

Results in the Paper

The results of Market1501, DukeMTMC-reID, MSMT17, P-DukeMTMC-reID, and Occluded DukeMTMC are provided below.

Model	Rank-1@Market1501	Rank-1@DukeMTMC-reID	Rank-1@MSMT17	Rank-1@P-DukeMTMC-reID	Rank-1@Occluded DukeMTMC
LDS	95.84	91.56	86.54	91.96	64.39

You can download these models in this link(code: huv8) and put them in the WEIGHT directory of the yml file. Then use the command below to evaluate them.

$ python projects\LDS\train_net.py --config-file projects/LDS/configs/Market1501/LDS_3Branch_mutual_test.yml

Acknowledgement

This repository is built upon the repository fast-reid.

Citation

If you find this project useful for your research, please kindly cite:

@article{zang2021learning,
	author = {Xianghao Zang and Ge Li and Wei Gao and Xiujun Shu},
	title = {Learning to disentangle scenes for person re-identification},
	journal = {Image and Vision Computing},
	volume = {116},
	pages = {104330},
	year = {2021},
	issn = {0262-8856},
	doi = {10.1016/j.imavis.2021.104330}
}

License

This repository is released under the GPL-2.0 License as found in the LICENSE file.