Towards better Validity: Dispersion based Clustering for unsupervised Person Re-identification

This respository is an official implementation of our paper titled "Towards better Validity: Dispersion based Clustering for unsupervised Person Re-identification". Click here to access the manuscript.

This code is based on the Open-ReID library and adopted from BUC.

Preparation

Dependencies

• Python 3.6
• PyTorch (version >= 0.4.1)
• h5py, scikit-learn, metric-learn, tqdm

Download datasets

• DukeMTMC-VideoReID: [Direct Link] [Google Drive] [BaiduYun]. This page contains more details and baseline code.
• MARS: [Google Drive] [BaiduYun].
• Market-1501: [Direct Link]
• DukeMTMC-reID: [Direct Link]
• Move the downloaded zip files to ./data/ and unzip here.

Usage

sh ./run.sh

--size_penalty parameter lambda to balance the intra-dispersion regularization term.

--merge_percent percent of data to merge at each iteration.

Citation

If you use this code or part of it in your work, please cite our paper:

@article{ding2019towards,
  title={Towards better Validity: Dispersion based Clustering for Unsupervised Person Re-identification},
  author={Ding, Guodong and Khan, Salman and Tang, Zhenmin and Zhang, Jian and Porikli, Fatih},
  journal={arXiv preprint arXiv:1906.01308},
  year={2019}
}