Balancing Discriminability and Transferability
for Source-Free Domain Adaptation
ICML 2022
Jogendra Nath Kundu*1 Akshay Kulkarni*1 Suvaansh Bhambri*1 Deepesh Mehta1 Shreyas Kulkarni1
Varun Jampani2 R. Venkatesh Babu1
1Video Analytics Lab, Indian Institute of Science 2Google Research
We analyze Source-Free Domain Adaptation (SFDA) from the perspective of the tradeoff between Discriminability and Transferability, and show that mixup between original and generic-domain translated samples yields an improved tradeoff, which improves the adaptation.
Video
Citation
If you find our work helpful in your research, please cite our work:
@inproceedings{kundu2022balancing,
title = {Balancing Discriminability and Transferability for Source-Free Domain Adaptation},
author = {Kundu, Jogendra Nath and Kulkarni, Akshay R and Bhambri, Suvaansh and Mehta, Deepesh and Kulkarni, Shreyas Anand and Jampani, Varun and Radhakrishnan, Venkatesh Babu},
booktitle = {Proceedings of the 39th International Conference on Machine Learning},
pages = {11710--11728},
year = {2022},
volume = {162},
series = {Proceedings of Machine Learning Research},
month = {17--23 Jul},
publisher = {PMLR}
}
License
This project is licenced under an [MIT License].
Contact
If you have any queries, please get in touch via email : jogendranathkundu@gmail.com