Drop to Adapt: Learning Discriminative Features for Unsupervised Domain Adaptation

TL;DR

Drop to Adapt (DTA) introduces adversarial dropout to enhance discriminative feature learning, improving domain adaptation performance in image classification and segmentation by enforcing the cluster assumption.

Contribution

The paper proposes Drop to Adapt, a novel adversarial dropout method that improves domain adaptation by learning more discriminative features aligned with the cluster assumption.

Findings

Achieves consistent improvements in image classification.

Enhances semantic segmentation performance.

Demonstrates efficacy across multiple experiments.

Abstract

Recent works on domain adaptation exploit adversarial training to obtain domain-invariant feature representations from the joint learning of feature extractor and domain discriminator networks. However, domain adversarial methods render suboptimal performances since they attempt to match the distributions among the domains without considering the task at hand. We propose Drop to Adapt (DTA), which leverages adversarial dropout to learn strongly discriminative features by enforcing the cluster assumption. Accordingly, we design objective functions to support robust domain adaptation. We demonstrate efficacy of the proposed method on various experiments and achieve consistent improvements in both image classification and semantic segmentation tasks. Our source code is available at https://github.com/postBG/DTA.pytorch.