DiDA: Disentangled Synthesis for Domain Adaptation

TL;DR

This paper introduces DiDA, a method that enhances unsupervised domain adaptation by disentangling common and domain-specific features, enabling better data synthesis and iterative performance improvements across models.

Contribution

It proposes a novel disentanglement-based approach that iteratively improves domain adaptation by synthesizing target data and refining feature separation.

Findings

Iterative disentanglement and adaptation improve performance.

Disentangled synthesis boosts domain adaptation accuracy.

Method is effective across various backbone models.

Abstract

Unsupervised domain adaptation aims at learning a shared model for two related, but not identical, domains by leveraging supervision from a source domain to an unsupervised target domain. A number of effective domain adaptation approaches rely on the ability to extract discriminative, yet domain-invariant, latent factors which are common to both domains. Extracting latent commonality is also useful for disentanglement analysis, enabling separation between the common and the domain-specific features of both domains. In this paper, we present a method for boosting domain adaptation performance by leveraging disentanglement analysis. The key idea is that by learning to separately extract both the common and the domain-specific features, one can synthesize more target domain data with supervision, thereby boosting the domain adaptation performance. Better common feature extraction, in turn, helps further improve the disentanglement analysis and disentangled synthesis. We show that iterating between domain adaptation and disentanglement analysis can consistently improve each other on several unsupervised domain adaptation tasks, for various domain adaptation backbone models.