Built-in Elastic Transformations for Improved Robustness

TL;DR

This paper introduces elastically-augmented convolutions (EAConv), a novel method that enhances CNN robustness against natural perturbations like elastic deformations, occlusions, and noise, improving performance on both clean and perturbed images.

Contribution

The paper proposes EAConv, a new convolutional approach that incorporates elastic perturbation bases, enabling CNNs to better handle natural view-point changes and perturbations.

Findings

Improved robustness on CIFAR-10 and STL-10 datasets.

Enhanced performance on clean images without data augmentation.

Better handling of occlusions, zoom, rotation, and Gaussian noise.

Abstract

We focus on building robustness in the convolutions of neural visual classifiers, especially against natural perturbations like elastic deformations, occlusions and Gaussian noise. Existing CNNs show outstanding performance on clean images, but fail to tackle naturally occurring perturbations. In this paper, we start from elastic perturbations, which approximate (local) view-point changes of the object. We present elastically-augmented convolutions (EAConv) by parameterizing filters as a combination of fixed elastically-perturbed bases functions and trainable weights for the purpose of integrating unseen viewpoints in the CNN. We show on CIFAR-10 and STL-10 datasets that the general robustness of our method on unseen occlusion, zoom, rotation, image cut and Gaussian perturbations improves, while significantly improving the performance on clean images without any data augmentation.