Safe Self-Refinement for Transformer-based Domain Adaptation

TL;DR

This paper introduces SSRT, a transformer-based approach for unsupervised domain adaptation that employs safe self-refinement to enhance transferability and reduce model collapse, achieving state-of-the-art results on multiple benchmarks.

Contribution

The paper proposes a novel transformer-based UDA method with a safe self-refinement strategy, improving transfer performance and robustness over existing CNN-based methods.

Findings

Transformer backbone surpasses CNN on DomainNet.

Safe self-refinement reduces model collapse.

Achieves top performance on multiple benchmarks.

Abstract

Unsupervised Domain Adaptation (UDA) aims to leverage a label-rich source domain to solve tasks on a related unlabeled target domain. It is a challenging problem especially when a large domain gap lies between the source and target domains. In this paper we propose a novel solution named SSRT (Safe Self-Refinement for Transformer-based domain adaptation), which brings improvement from two aspects. First, encouraged by the success of vision transformers in various vision tasks, we arm SSRT with a transformer backbone. We find that the combination of vision transformer with simple adversarial adaptation surpasses best reported Convolutional Neural Network (CNN)-based results on the challenging DomainNet benchmark, showing its strong transferable feature representation. Second, to reduce the risk of model collapse and improve the effectiveness of knowledge transfer between domains with large gaps, we propose a Safe Self-Refinement strategy. Specifically, SSRT utilizes predictions of perturbed target domain data to refine the model. Since the model capacity of vision transformer is large and predictions in such challenging tasks can be noisy, a safe training mechanism is designed to adaptively adjust learning configuration. Extensive evaluations are conducted on several widely tested UDA benchmarks and SSRT achieves consistently the best performances, including 85.43% on Office-Home, 88.76% on VisDA-2017 and 45.2% on DomainNet.