Consistency Regularization for Generalizable Source-free Domain Adaptation

TL;DR

This paper introduces a consistency regularization framework for source-free domain adaptation that improves model generalization to unseen test data by using pseudo-labels, data augmentation, and global calibration techniques.

Contribution

It proposes a novel consistency regularization approach with pseudo-label sampling and global calibration to enhance generalization in source-free domain adaptation.

Findings

Achieves state-of-the-art results on SFDA benchmarks.

Demonstrates robustness on unseen testing datasets.

Improves generalization over existing methods.

Abstract

Source-free domain adaptation (SFDA) aims to adapt a well-trained source model to an unlabelled target domain without accessing the source dataset, making it applicable in a variety of real-world scenarios. Existing SFDA methods ONLY assess their adapted models on the target training set, neglecting the data from unseen but identically distributed testing sets. This oversight leads to overfitting issues and constrains the model's generalization ability. In this paper, we propose a consistency regularization framework to develop a more generalizable SFDA method, which simultaneously boosts model performance on both target training and testing datasets. Our method leverages soft pseudo-labels generated from weakly augmented images to supervise strongly augmented images, facilitating the model training process and enhancing the generalization ability of the adapted model. To leverage more potentially useful supervision, we present a sampling-based pseudo-label selection strategy, taking samples with severer domain shift into consideration. Moreover, global-oriented calibration methods are introduced to exploit global class distribution and feature cluster information, further improving the adaptation process. Extensive experiments demonstrate our method achieves state-of-the-art performance on several SFDA benchmarks, and exhibits robustness on unseen testing datasets.