Awesome

Learning Modal-Invariant and Temporal-Memory for Video-based Visible-Infrared Person Re-Identification

This is the page for HITSZ-VCM dataset.

HITSZ-VCM is a large scale video-based cross-modal dataset for person re-identification.

Usage

• Download the Train data, Test data and info from Google Drive or from Baidu Netdisk (password: kxzs).

• Unzip the data and put the data in the same directory. The path to the data is placed in data_manager.py.

• Data structure.

    |____ data/
         |____ 0001/
              |____ ir/
              |____ rgb/
         |____ 0002/
         …
         |____ 0927/
         |____test_name.txt
         |____track_test_info.txt
         |____query_IDX.txt
         |____train_name.txt
         |____track_train_info.txt
     
  

License

• Data is available for academic only. This means that it cannot be used to train models for commercial use.
• You may NOT redistribute the dataset. This includes posting it on a website or sending it to others.
• Models trained using our data may only be distributed (posted on the internet or given to others) under the condition that the model can only be used for non-commercial uses.
• Any publications utilizing this dataset should reference our paper.

By downloading or using any of these files, you agree to be bound by and comply with the license agreement.

@inproceedings{lin2022learning,
  title={Learning Modal-Invariant and Temporal-Memory for Video-Based Visible-Infrared Person Re-Identification},
  author={Lin, Xinyu and Li, Jinxing and Ma, Zeyu and Li, Huafeng and Li, Shuang and Xu, Kaixiong and Lu, Guangming and Zhang, David},
  booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
  pages={20973--20982},
  year={2022}
}