HybridAugment++: Unified Frequency Spectra Perturbations for Model Robustness
This addresses robustness issues in CNNs for computer vision applications, but it is incremental as it builds on existing frequency-centric approaches.
The paper tackles the problem of poor generalization in CNNs under distribution shifts by proposing HybridAugment and HybridAugment++, data augmentation methods that reduce reliance on high-frequency and amplitude components, achieving competitive or better results on clean accuracy, corruption benchmarks, adversarial robustness, and out-of-distribution detection across multiple datasets.
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.