A Focally Discriminative Loss for Unsupervised Domain Adaptation

A PyTorch implementation of 'A Focally Discriminative Loss for Unsupervised Domain Adaptation' which has published on The 28th International Conference on Neural Information Processing

Requirement

• python 3
• pytorch 1.0

Usage

1. You can download Office31 dataset here. And then unrar dataset in ./dataset/.
2. You can change the src and tgt in main.py to set different transfer tasks.
3. Run python main.py.

Note that for tasks D-A and W-A, setting epochs = 800 or larger could achieve better performance.

Reference

@inproceedings{sun2021focally,
  title={A focally discriminative loss for unsupervised domain adaptation},
  author={Sun, Dongting and Wang, Mengzhu and Ma, Xurui and Zhang, Tianming and Yin, Nan and Yu, Wei and Luo, Zhigang},
  booktitle={International Conference on Neural Information Processing},
  pages={54--64},
  year={2021},
  organization={Springer}
}