Learning Intra and Inter-Camera Invariance for Isolated Camera supervised Person Re-Identification

[Paper]

This repository is the implementation of Learning Intra and Inter-Camera Invariance for Isolated Camera supervised Person Re-Identification, ACM MultiMedia 2023. The proposed method IICI targets at isolated camera supervised re-ID, and achieves state-of-the-art performance on multiple re-ID benchmarks.

Requirements

Environment

PyTorch >= 1.8

Installation

git clone https://github.com/Terminator8758/IICI.git
cd IICI

Prepare Datasets

Download the re-ID datasets Market-1501, MSMT17. Then put them under a folder such as '/path/to/dataset/'.

Training

We utilize 4 GPUs for training. Performance reported in the paper can be obtained by running the following commands:

Train on Market-1501 using ResNet-Nonlocal backbone (default):

bash train_market.sh 

Train on Market-1501 using ViT-S backbone:

bash train_market_vit.sh 

Train on MSMT17 using ResNet-Nonlocal backbone (default):

bash train_msmt.sh 

Train on MSMT17 using ViT-S backbone:

bash train_msmt_vit.sh 

Result

Citation

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

@article{2023_wang_iici,
    title={Learning Intra and Inter-Camera Invariance for Isolated Camera supervised Person Re-Identification},
    author={Menglin Wang and Xiaojin Gong},
    journal={ACM MultiMedia Conference},
    year={2023}
}