HybridAugment++: Unified Frequency Spectra Perturbations for Model Robustness

2023-10-06
Yücel, Mehmet Kerim
Cinbiş, Ramazan Gökberk
DUYGULU ŞAHİN, PINAR
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 high-light 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 CIFAR100-C), adversarial robustness on CIFAR-10 and out-ofdistribution 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(1).
IEEE/CVF International Conference on Computer Vision (ICCV)
Citation Formats
M. K. Yücel, R. G. Cinbiş, and P. DUYGULU ŞAHİN, “HybridAugment++: Unified Frequency Spectra Perturbations for Model Robustness,” presented at the IEEE/CVF International Conference on Computer Vision (ICCV), Fransa, 2023, Accessed: 00, 2024. [Online]. Available: https://openaccess.thecvf.com/content/ICCV2023/papers/Yucel_HybridAugment_Unified_Frequency_Spectra_Perturbations_for_Model_Robustness_ICCV_2023_paper.pdf.