HybridAugment++: Unified Frequency Spectra Perturbations for Model Robustness

TL;DR

HybridAugment++ is a hierarchical data augmentation technique that improves CNN robustness against distribution shifts by unifying frequency spectrum perturbations, focusing on phase information, and maintaining high accuracy on clean data.

Contribution

It introduces HybridAugment++, a novel hierarchical augmentation method unifying frequency spectrum perturbations to enhance CNN robustness and generalization.

Findings

Outperforms state-of-the-art on clean accuracy benchmarks

Improves robustness on corruption and adversarial tests

Requires minimal code and no extra data or models

Abstract

Convolutional Neural Networks (CNN) are known to exhibit poor generalization performance under distribution shifts. Their generalization have been studied extensively, and one line of work approaches the problem from a frequency-centric perspective. These studies highlight the fact that humans and CNNs might focus on different frequency components of an image. First, inspired by these observations, we propose a simple yet effective data augmentation method HybridAugment that reduces the reliance of CNNs on high-frequency components, and thus improves their robustness while keeping their clean accuracy high. Second, we propose HybridAugment++, which is a hierarchical augmentation method that attempts to unify various frequency-spectrum augmentations. HybridAugment++ builds on HybridAugment, and also reduces the reliance of CNNs on the amplitude component of images, and promotes phase information instead. This unification results in competitive to or better than state-of-the-art results on clean accuracy (CIFAR-10/100 and ImageNet), corruption benchmarks (ImageNet-C, CIFAR-10-C and CIFAR-100-C), adversarial robustness on CIFAR-10 and out-of-distribution detection on various datasets. HybridAugment and HybridAugment++ are implemented in a few lines of code, does not require extra data, ensemble models or additional networks.