Improving Robustness Without Sacrificing Accuracy with Patch Gaussian Augmentation

TL;DR

Patch Gaussian augmentation enhances model robustness to corruptions and improves accuracy on clean data by adding noise to image patches, overcoming the traditional robustness-accuracy trade-off.

Contribution

Introducing Patch Gaussian, a simple patch-based noise augmentation that improves robustness and accuracy simultaneously, compatible with existing regularization and augmentation methods.

Findings

Achieves state-of-the-art results on CIFAR-10 and ImageNet corruption benchmarks.

Reduces sensitivity to high-frequency noise while preserving relevant high-frequency information.

Enhances performance on COCO object detection when combined with other methods.

Abstract

Deploying machine learning systems in the real world requires both high accuracy on clean data and robustness to naturally occurring corruptions. While architectural advances have led to improved accuracy, building robust models remains challenging. Prior work has argued that there is an inherent trade-off between robustness and accuracy, which is exemplified by standard data augment techniques such as Cutout, which improves clean accuracy but not robustness, and additive Gaussian noise, which improves robustness but hurts accuracy. To overcome this trade-off, we introduce Patch Gaussian, a simple augmentation scheme that adds noise to randomly selected patches in an input image. Models trained with Patch Gaussian achieve state of the art on the CIFAR-10 and ImageNetCommon Corruptions benchmarks while also improving accuracy on clean data. We find that this augmentation leads to reduced sensitivity to high frequency noise(similar to Gaussian) while retaining the ability to take advantage of relevant high frequency information in the image (similar to Cutout). Finally, we show that Patch Gaussian can be used in conjunction with other regularization methods and data augmentation policies such as AutoAugment, and improves performance on the COCO object detection benchmark.