Self-supervised Geometric Features Discovery with Interpretable Attention for Vehicle Re-Identification and Beyond

Introduction

This is the code of our ICCV21 paper.

Datasets

• veri776
• vehicleID

Tutorial

train

Input arguments for the training scripts are unified in args.py.

python train.py

test

Use --evaluate to switch to the evaluation mode.

BibTeX

If you use this code in your project, please cite our paper:

@inproceedings{li2021self,
  title={Self-Supervised Geometric Features Discovery via Interpretable Attention for Vehicle Re-Identification and Beyond},
  author={Li, Ming and Huang, Xinming and Zhang, Ziming},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={194--204},
  year={2021}
}

Thanks

Our code refers to ReID strong baseline and D2Net.