Gpn-star/SCNet

Pytorch implementation of 'Semantic-aware Consistency Network for Cloth-changing Person Re-Identification. In ACM MM, 2023.'

Semantic-aware Consistency Network for Cloth-changing Person Re-Identification (ACM MM, 2023) [Paper]

Overall Framework

Requirements

• Python 3.6
• Pytorch 1.6.0
• yacs
• apex
• GeForce RTX 4090 × 2

Get Started

• Clone this repo:

  git clone https://github.com/Gpn-star/SCNet.git
  cd SCNet

• Download datasets:
  LTCC: [Offical Link]
  PRCC: [Offical Link]
  Vc-Clothes: [Offical Link]
  DeepChange: [Offical Link]

• Download human parsing results:
  LTCC: [Google Drive]
  PRCC: [Google Drive]
  Vc-Clothes: [Google Drive]
  DeepChange: [Google Drive]

• Arrange datasets according to the following structure：

Dataset/
├── LTCC_ReID/
│   ├── ...
│   └── processed
├── PRCC/
|   ├── rgb / processed
│   └── sketch
├── Vc-Clothes/
|   ├── ...
|   └── processed
└── DeepChange/
    ├── ...
    └── processed

• Replace _C.DATA.ROOT and _C.OUTPUT in configs/default_img.pywith your own data root path and output path, respectively.

• Run script.sh

Citation

If you find this code useful for your research, please cite our paper:

@inproceedings{guo2023semantic,
  title={Semantic-aware Consistency Network for Cloth-changing Person Re-Identification},
  author={Guo, Peini and Liu, Hong and Wu, Jianbing and Wang, Guoquan and Wang, Tao},
  booktitle={Proceedings of the 31st ACM International Conference on Multimedia},
  pages={8730--8739},
  year={2023}
}