Universal Domain Adaptation through Self Supervision

TL;DR

This paper introduces DANCE, a universal domain adaptation framework that employs self-supervised neighborhood clustering and entropy-based feature alignment to handle arbitrary category shifts across domains.

Contribution

DANCE is a novel framework that combines neighborhood clustering and entropy optimization to adapt to various domain shifts without prior setting knowledge.

Findings

DANCE outperforms existing methods in open-set, open-partial, and partial domain adaptation.

The neighborhood clustering technique effectively learns target domain structure.

Entropy-based alignment improves feature transferability.

Abstract

Unsupervised domain adaptation methods traditionally assume that all source categories are present in the target domain. In practice, little may be known about the category overlap between the two domains. While some methods address target settings with either partial or open-set categories, they assume that the particular setting is known a priori. We propose a more universally applicable domain adaptation framework that can handle arbitrary category shift, called Domain Adaptative Neighborhood Clustering via Entropy optimization (DANCE). DANCE combines two novel ideas: First, as we cannot fully rely on source categories to learn features discriminative for the target, we propose a novel neighborhood clustering technique to learn the structure of the target domain in a self-supervised way. Second, we use entropy-based feature alignment and rejection to align target features with the source, or reject them as unknown categories based on their entropy. We show through extensive experiments that DANCE outperforms baselines across open-set, open-partial and partial domain adaptation settings. Implementation is available at https://github.com/VisionLearningGroup/DANCE.