Celeb-reID

This repository contains Datasets and Code for our paper Beyond Scalar Neuron: Adopting Vector-Neuron Capsules for Long-Term Person Re-Identification and Celebrities-ReID: A Benchmark for Clothes Variation in Long-Term Person Re-Identification.

1.Dataset

You can directly download our datasets from OneDrive Celeb-reID and its light version Celeb-reID-light.

Baidu Cloud Link:

For Celeb-reID: Link：https://pan.baidu.com/s/1sKkO4l8FyzM7iXnzjPyWHQ code：ix2j

For Celeb-reID-light: Link：https://pan.baidu.com/s/13DSJ1PK_AEF9TEGi30eIQA code：14k5

The data split of Celeb-reID is as follows:

split	training	query	gallery	total
#ID	632	420	420	1,052
#Images	20,208	2,972	11,006	34,186

The data split of Celeb-reID-light is as follows:

split	training	query	gallery	total
#ID	490	100	100	590
#Images	9021	887	934	10,842

Note: The two datasets should be used for research only. Please DO NOT distribute or use it for commercial purpose.

2.Code

Our code (ReIDCaps) is implemented by Pytorch(>=1.0.0) and python(anaconda 3.6) with Ubuntu.

To run the training code:

First: Download the Celeb-reID dataset to your own path. Do body part partition by copying the part_partition.m file to the path of your dataset. Changing name=gallery, name=query, and name=train to get the body part partition files in the path of your Celeb-reID dataset.

Second: Directly run the run_train.sh by bash ./run_train.sh on console. You may change the path of dataset in the file of run_train.sh according to your own path.

To run the testing code:

Just simply run the 'test.py' file. You may change the path of dataset and logs(the path of trained model) according to your own path.

You can also directly download our trained_model to get the perfomance reported in our paper Beyond Scalar Neuron: Adopting Vector-Neuron Capsules for Long-Term Person Re-Identification, TCSVT2019(see Table VII ReIDCaps*(ours)).

Method	mAP	rank-1	rank-5
ReIDCaps*(Ours)	15.8%	63.0%	76.3%

Citation

Please cite this paper in your publications if it helps your research:

@article{huang2019beyond,
  title={Beyond Scalar Neuron: Adopting Vector-Neuron Capsules for Long-Term Person Re-Identification},
  author={Huang, Yan and Xu, Jingsong and Wu, Qiang and Zhong, Yi and Zhang, Peng and Zhang, Zhaoxiang},
  journal={Transactions on Circuits and Systems for Video Technology (TCSVT)},
  year={2019},
  publisher={IEEE}
}

@inproceedings{huang2021clothing,
  title={Clothing status awareness for long-term person re-identification},
  author={Huang, Yan and Wu, Qiang and Xu, JingSong and Zhong, Yi and Zhang, ZhaoXiang},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={11895--11904},
  year={2021}
}

@inproceedings{huang2019celebrities,
  title={Celebrities-ReID: A Benchmark for Clothes Variation in Long-Term Person Re-Identification},
  author={Huang, Yan and Wu, Qiang and Xu, Jingsong and Zhong, Yi},
  booktitle={International Joint Conference on Neural Networks (IJCNN)},
  pages={1--8},
  year={2019},
  organization={IEEE}
}