FixBi: Bridging Domain Spaces for Unsupervised Domain Adaptation
This work aims to improve the performance of unsupervised domain adaptation, particularly for scenarios with large domain discrepancies, which is a common challenge for practitioners applying models across different datasets.
This paper addresses large domain discrepancies in Unsupervised Domain Adaptation (UDA) by augmenting intermediate domains between source and target domains using a fixed ratio-based mixup. The method trains complementary source-dominant and target-dominant models, which then learn from each other through confidence-based bidirectional matching and self-penalization, gradually transferring domain knowledge.
Unsupervised domain adaptation (UDA) methods for learning domain invariant representations have achieved remarkable progress. However, most of the studies were based on direct adaptation from the source domain to the target domain and have suffered from large domain discrepancies. In this paper, we propose a UDA method that effectively handles such large domain discrepancies. We introduce a fixed ratio-based mixup to augment multiple intermediate domains between the source and target domain. From the augmented-domains, we train the source-dominant model and the target-dominant model that have complementary characteristics. Using our confidence-based learning methodologies, e.g., bidirectional matching with high-confidence predictions and self-penalization using low-confidence predictions, the models can learn from each other or from its own results. Through our proposed methods, the models gradually transfer domain knowledge from the source to the target domain. Extensive experiments demonstrate the superiority of our proposed method on three public benchmarks: Office-31, Office-Home, and VisDA-2017.