SDA-VI-ReID

Overview

This repository contains the PyTorch implementation of SDA-VI-ReID, a novel approach to Visible-Infrared Person Re-Identification (VI-ReID). Our method addresses the challenging problem of matching individuals across visible and infrared image modalities.

Training and Testing

Sample scripts for training and testing across various datasets are provided in run.sh.

To change the number of target IDs, modify the --target_ids argument:

python train.py --target_ids 10

Acknowledgments

Our code builds upon and extends the following excellent works:

• MMD-ReID
• AGW

Please refer to the official repositories for additional details on:

• Dataset downloading
• Data preparation
• Baseline hyperparameter selection

Citation

If you find our work useful in your research, please consider citing:

@inproceedings{sahu2024sda,
  title={Supervised Domain Adaptation for Data-Efficient Visible-Infrared Person Re-identification},
  author={Sahu, Mihir and Singh, Arjun and Kolekar, Maheshkumar},
  booktitle={Pattern Recognition},
  year={2025}
}