Challenges of Adversarial Image Augmentations

TL;DR

This paper investigates the effectiveness of adversarial image augmentation strategies, comparing their performance to random and curriculum-based approaches, and suggests stochasticity plays a key role in their success.

Contribution

The study demonstrates that random augmentations remain competitive with adversarial methods and proposes that stochasticity in policy controllers contributes to their effectiveness.

Findings

Random augmentations are still competitive with adversarial approaches.

Stochasticity in augmentation policies introduces a beneficial curriculum effect.

Adversarial AutoAugment's success may be due to policy stochasticity rather than adversarial training alone.

Abstract

Image augmentations applied during training are crucial for the generalization performance of image classifiers. Therefore, a large body of research has focused on finding the optimal augmentation policy for a given task. Yet, RandAugment [2], a simple random augmentation policy, has recently been shown to outperform existing sophisticated policies. Only Adversarial AutoAugment (AdvAA) [11], an approach based on the idea of adversarial training, has shown to be better than RandAugment. In this paper, we show that random augmentations are still competitive compared to an optimal adversarial approach, as well as to simple curricula, and conjecture that the success of AdvAA is due to the stochasticity of the policy controller network, which introduces a mild form of curriculum.