Spectral Adversarial MixUp for Few-Shot Unsupervised Domain Adaptation

TL;DR

This paper introduces a novel spectral adversarial MixUp technique for few-shot unsupervised domain adaptation, improving model generalization with limited unlabeled target data by characterizing and suppressing spectral sensitivities.

Contribution

The paper proposes a spectral sensitivity map and a Sensitivity-guided Spectral Adversarial MixUp method for effective domain adaptation with few target samples.

Findings

Enhanced model generalization in target domain

Effective suppression of spectral sensitivities

Improved performance on multiple datasets

Abstract

Domain shift is a common problem in clinical applications, where the training images (source domain) and the test images (target domain) are under different distributions. Unsupervised Domain Adaptation (UDA) techniques have been proposed to adapt models trained in the source domain to the target domain. However, those methods require a large number of images from the target domain for model training. In this paper, we propose a novel method for Few-Shot Unsupervised Domain Adaptation (FSUDA), where only a limited number of unlabeled target domain samples are available for training. To accomplish this challenging task, first, a spectral sensitivity map is introduced to characterize the generalization weaknesses of models in the frequency domain. We then developed a Sensitivity-guided Spectral Adversarial MixUp (SAMix) method to generate target-style images to effectively suppresses the model sensitivity, which leads to improved model generalizability in the target domain. We demonstrated the proposed method and rigorously evaluated its performance on multiple tasks using several public datasets.