Official implementation of "Color Prompting for Data-Free Continual Unsupervised Domain Adaptive Person Re-Identification".
- We propose a Color Prompting (CoP) method for for data-free continual unsupervised domain adaptive person Re-ID.
- CoP employs a color style transfer process to rehearse the information of previous tasks without storing the data.
- The proposed CoP achieves 6.7% and 8.1% average rank-1 improvements over the replay method on seen and unseen domains, respectively.
Put the datasets into ../data
.
Pre-train the model on the source domain.
sh scripts/pretrain_market.sh
Conduct single domain adaptation.
sh scripts/single_continual.sh
Conduct multiple domain adpatation.
sh scripts/multi_continual.sh
If you find this work helpful, please cite:
@article{gu2023color,
title={Color Prompting for Data-Free Continual Unsupervised Domain Adaptive Person Re-Identification},
author={Gu, Jianyang and Luo, Hao and Wang, Kai and Jiang, Wei and You, Yang and Zhao, Jian},
journal={arXiv preprint arXiv:2308.10716},
year={2023}
}