A simple way to make neural networks robust against diverse image corruptions
This addresses robustness issues in image recognition models for real-world applications, offering an incremental improvement over existing methods.
The paper tackled the problem of neural networks' performance degradation on unseen image corruptions by demonstrating that simple training with additive Gaussian and Speckle noise achieves state-of-the-art results on benchmarks like ImageNet-C and MNIST-C, and further improves with adversarial training against worst-case noise.
The human visual system is remarkably robust against a wide range of naturally occurring variations and corruptions like rain or snow. In contrast, the performance of modern image recognition models strongly degrades when evaluated on previously unseen corruptions. Here, we demonstrate that a simple but properly tuned training with additive Gaussian and Speckle noise generalizes surprisingly well to unseen corruptions, easily reaching the previous state of the art on the corruption benchmark ImageNet-C (with ResNet50) and on MNIST-C. We build on top of these strong baseline results and show that an adversarial training of the recognition model against uncorrelated worst-case noise distributions leads to an additional increase in performance. This regularization can be combined with previously proposed defense methods for further improvement.