Domain Generalization Emerges from Dreaming

TL;DR

This paper introduces Stylized Dream, a novel data augmentation method that reduces texture bias in deep neural networks by encouraging shape-based representations, thereby improving out-of-distribution generalization.

Contribution

The paper proposes a new optimization-based augmentation using AdaIN and regularization to promote shape bias, achieving state-of-the-art domain generalization results.

Findings

Outperforms existing methods on multiple benchmarks

Reduces texture bias in DNNs effectively

Improves out-of-distribution generalization

Abstract

Recent studies have proven that DNNs, unlike human vision, tend to exploit texture information rather than shape. Such texture bias is one of the factors for the poor generalization performance of DNNs. We observe that the texture bias negatively affects not only in-domain generalization but also out-of-distribution generalization, i.e., Domain Generalization. Motivated by the observation, we propose a new framework to reduce the texture bias of a model by a novel optimization-based data augmentation, dubbed Stylized Dream. Our framework utilizes adaptive instance normalization (AdaIN) to augment the style of an original image yet preserve the content. We then adopt a regularization loss to predict consistent outputs between Stylized Dream and original images, which encourages the model to learn shape-based representations. Extensive experiments show that the proposed method achieves state-of-the-art performance in out-of-distribution settings on public benchmark datasets: PACS, VLCS, OfficeHome, TerraIncognita, and DomainNet.