Do CNNs Encode Data Augmentations?

TL;DR

This paper investigates whether CNN features encode data augmentation transformations, finding that early layers encode these features strongly, but the signal diminishes in deeper layers, highlighting the layered encoding of transformations.

Contribution

Introduces a systematic method to analyze which CNN layers encode data augmentation features, revealing early layers capture these transformations more effectively.

Findings

Early CNN layers encode augmentation features strongly.

The augmentation signal diminishes in deeper layers.

Neural network features can predict various data transformations with high accuracy.

Abstract

Data augmentations are important ingredients in the recipe for training robust neural networks, especially in computer vision. A fundamental question is whether neural network features encode data augmentation transformations. To answer this question, we introduce a systematic approach to investigate which layers of neural networks are the most predictive of augmentation transformations. Our approach uses features in pre-trained vision models with minimal additional processing to predict common properties transformed by augmentation (scale, aspect ratio, hue, saturation, contrast, and brightness). Surprisingly, neural network features not only predict data augmentation transformations, but they predict many transformations with high accuracy. After validating that neural networks encode features corresponding to augmentation transformations, we show that these features are encoded in the early layers of modern CNNs, though the augmentation signal fades in deeper layers.