/DaC

Code for Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning (NeurIPS 2022).

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

Official implementation for DaC (NeurIPS 2022)

Code for our NeurIPS 2022 paper 'Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning.' Paper (openreview)

Dataset:

Please manually download the datasets Office-Home, VisDA-C, DomainNet from the official websites, and denote the path of images in each '.txt' (best under folder /DATANAME/data/).

VisDA Training:

 cd VisDA/
 sh run_source
 sh run_train

Citation

We would be grateful if you cite this paper:

@inproceedings{zhang2022dac, 
 title={Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning.}, 
 author={Zhang, Ziyi and Chen, Weikai and Cheng, Hui and Li, Zhen and Li, Siyuan and Lin, Liang and Li, Guanbin}, 
 booktitle={Conference on Neural Information Processing Systems (NeurIPS)},  
 year={2022}
}

Contact

Acknowledgement

This code is based on SHOT