Robust Pseudo-label Learning with Neighbor Relation for Unsupervised Visible-Infrared Person Re-Identification

Xiangbo Yin* · Jiangming Shi* · Yachao Zhang · Yang Lu · Zhizhong Zhang · Yuan Xie† · Yanyun Qu†

Paper Link

This is an official code implementation of "Robust Pseudo-label Learning with Neighbor Relation for Unsupervised Visible-Infrared Person Re-Identification", Which is accepted by ACMMM 2024.

Requirements

• python 3.8.13
• torch 1.8.0
• torchvision 0.9.0
• scikit-learn 1.2.2
• POT 0.9.3

Dataset Preprocessing

python prepare_sysu.py   # for SYSU-MM01
python prepare_regdb.py  # for RegDB

You need to change the dataset path to your own path in the prepare_sysu.py and prepare_regdb.py.

Training

sh run_train_sysu.sh     # for SYSU-MM01
sh run_train_regdb.sh    # for RegDB

You need to download the baseline for SYSU-MM01 dataset and put it in baseline/sysu_s1/.

Testing

sh test_sysu.sh          # for SYSU-MM01
sh test_regdb.sh         # for RegDB

Citation

If our work is helpful for your research, please consider citing:

@article{yin2024robust,
  title={Robust Pseudo-label Learning with Neighbor Relation for Unsupervised Visible-Infrared Person Re-Identification},
  author={Yin, Xiangbo and Shi, Jiangming and Zhang, Yachao and Lu, Yang and Zhang, Zhizhong and Xie, Yuan and Qu, Yanyun},
  journal={arXiv preprint arXiv:2405.05613},
  year={2024}
}

@article{shi2024multi,
  title={Multi-Memory Matching for Unsupervised Visible-Infrared Person Re-Identification},
  author={Shi, Jiangming and Yin, Xiangbo and Chen, Yeyun and Zhang, Yachao and Zhang, Zhizhong and Xie, Yuan and Qu, Yanyun},
  journal={arXiv preprint arXiv:2401.06825},
  year={2024}
}

@inproceedings{shi2023dpis,
  title={Dual pseudo-labels interactive self-training for semi-supervised visible-infrared person re-identification},
  author={Shi, Jiangming and Zhang, Yachao and Yin, Xiangbo and Xie, Yuan and Zhang, Zhizhong and Fan, Jianping and Shi, Zhongchao and Qu, Yanyun},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={11218--11228},
  year={2023}

Contact

[email protected]; [email protected].

Acknowledgements

This work is developed based on works of ADCA(ACMMM2022), PGM(CVPR2023), OTLA(ECCV2022). We sincerely thank all developers for their high-quality works.