Adversarial Bayesian Augmentation for Single-Source Domain Generalization

TL;DR

This paper introduces Adversarial Bayesian Augmentation (ABA), a novel method that generates diverse image augmentations using adversarial and Bayesian techniques to improve single-source domain generalization across various shifts.

Contribution

ABA is the first approach to combine adversarial learning with Bayesian neural networks for generating augmentations in single-source domain generalization.

Findings

ABA outperforms all previous state-of-the-art methods.

Effective across style, subpopulation, and medical imaging shifts.

Generates diverse augmentations that enhance generalization.

Abstract

Generalizing to unseen image domains is a challenging problem primarily due to the lack of diverse training data, inaccessible target data, and the large domain shift that may exist in many real-world settings. As such data augmentation is a critical component of domain generalization methods that seek to address this problem. We present Adversarial Bayesian Augmentation (ABA), a novel algorithm that learns to generate image augmentations in the challenging single-source domain generalization setting. ABA draws on the strengths of adversarial learning and Bayesian neural networks to guide the generation of diverse data augmentations -- these synthesized image domains aid the classifier in generalizing to unseen domains. We demonstrate the strength of ABA on several types of domain shift including style shift, subpopulation shift, and shift in the medical imaging setting. ABA outperforms all previous state-of-the-art methods, including pre-specified augmentations, pixel-based and convolutional-based augmentations.