Rethinking Domain Generalization Baselines

TL;DR

This paper examines style transfer data augmentation for domain generalization, revealing that current state-of-the-art methods often lose effectiveness when combined with this augmentation, highlighting the need for new approaches.

Contribution

It introduces a simple, inexpensive style transfer augmentation strategy and critically evaluates its impact on existing domain generalization methods.

Findings

Style transfer augmentation improves baseline robustness.

State-of-the-art methods lose effectiveness with augmentation.

Highlights need for new domain generalization techniques.

Abstract

Despite being very powerful in standard learning settings, deep learning models can be extremely brittle when deployed in scenarios different from those on which they were trained. Domain generalization methods investigate this problem and data augmentation strategies have shown to be helpful tools to increase data variability, supporting model robustness across domains. In our work we focus on style transfer data augmentation and we present how it can be implemented with a simple and inexpensive strategy to improve generalization. Moreover, we analyze the behavior of current state of the art domain generalization methods when integrated with this augmentation solution: our thorough experimental evaluation shows that their original effect almost always disappears with respect to the augmented baseline. This issue open new scenarios for domain generalization research, highlighting the need of novel methods properly able to take advantage of the introduced data variability.