UPA

This the official implementation for the paper "Uncertainty-Aware Pseudo-Label Filtering for Source-Free Unsupervised Domain Adaptation". Accepted to Neurocomputing.

Framework:

Usage

Environments and Datasets preparation please refer to SHOT.

Training

1. Train source models

    sh source.sh

2. Soure-free domain adaptation in target domains

    sh run_office.sh/run_office-home.sh/run_visda.sh/run_domainnet126.sh

Citation

If you find this code useful for your research, please cite our papers

    @article{chen_uncertainty-aware_2024,
	title = {Uncertainty-aware pseudo-label filtering for source-free unsupervised domain adaptation},
	volume = {575},
	issn = {0925-2312},
	pages = {127190},
	journaltitle = {Neurocomputing},
	author = {Chen, Xi and Yang, Haosen and Zhang, Huicong and Yao, Hongxun and Zhu, Xiatian},
	date = {2024},
}