Abstract: Generalization under distribution shifts has been a great challenge in computer vision. The prevailing practice of directly employing the one-hot labels as the training targets in domain ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback