WSAM: Visual Explanations from Style Augmentation as Adversarial Attacker and Their Influence in Image Classification

TL;DR

This paper introduces a stochastic style augmentation method that enhances CNN robustness against stylized images, surpasses previous techniques in performance, and provides insights into model interpretations under style variations.

Contribution

A novel stochastic style augmentation algorithm using noise addition improves model robustness and performance, with comprehensive analysis of interpretability under style variations.

Findings

Models show increased robustness against stylized images.

The proposed method outperforms previous style transfer techniques.

Analysis reveals how style variations influence model interpretations.

Abstract

Currently, style augmentation is capturing attention due to convolutional neural networks (CNN) being strongly biased toward recognizing textures rather than shapes. Most existing styling methods either perform a low-fidelity style transfer or a weak style representation in the embedding vector. This paper outlines a style augmentation algorithm using stochastic-based sampling with noise addition to improving randomization on a general linear transformation for style transfer. With our augmentation strategy, all models not only present incredible robustness against image stylizing but also outperform all previous methods and surpass the state-of-the-art performance for the STL-10 dataset. In addition, we present an analysis of the model interpretations under different style variations. At the same time, we compare comprehensive experiments demonstrating the performance when applied to deep neural architectures in training settings.