Randomized Adversarial Style Perturbations for Domain Generalization

TL;DR

This paper introduces RASP, a novel domain generalization method that perturbs feature styles adversarially to improve model robustness across unseen domains, enhanced by NFM to balance learning and robustness.

Contribution

The paper proposes RASP, a new style perturbation technique for domain generalization, combined with NFM to prevent degradation of source domain knowledge during training.

Findings

RASP improves domain generalization performance on multiple benchmarks.

NFM effectively balances learning original features and robustness.

The combined approach outperforms existing methods in large-scale benchmarks.

Abstract

We propose a novel domain generalization technique, referred to as Randomized Adversarial Style Perturbation (RASP), which is motivated by the observation that the characteristics of each domain are captured by the feature statistics corresponding to style. The proposed algorithm perturbs the style of a feature in an adversarial direction towards a randomly selected class, and makes the model learn against being misled by the unexpected styles observed in unseen target domains. While RASP is effective to handle domain shifts, its naive integration into the training procedure might degrade the capability of learning knowledge from source domains because it has no restriction on the perturbations of representations. This challenge is alleviated by Normalized Feature Mixup (NFM), which facilitates the learning of the original features while achieving robustness to perturbed representations via their mixup during training. We evaluate the proposed algorithm via extensive experiments on various benchmarks and show that our approach improves domain generalization performance, especially in large-scale benchmarks.