Data Augmentation Revisited: Rethinking the Distribution Gap between Clean and Augmented Data

TL;DR

This paper analyzes the impact of data augmentation on model generalization, revealing a trade-off between reduced generalization error and increased empirical risk, and proposes a simple refinement method to improve accuracy.

Contribution

It offers an analytical perspective on data augmentation as regularization, quantifies the distribution gap, and introduces a refinement approach for better performance.

Findings

Data augmentation reduces generalization error but slightly increases empirical risk.

Using less-augmented data for refinement improves model accuracy.

The approach improves performance on image classification and object detection benchmarks.

Abstract

Data augmentation has been widely applied as an effective methodology to improve generalization in particular when training deep neural networks. Recently, researchers proposed a few intensive data augmentation techniques, which indeed improved accuracy, yet we notice that these methods augment data have also caused a considerable gap between clean and augmented data. In this paper, we revisit this problem from an analytical perspective, for which we estimate the upper-bound of expected risk using two terms, namely, empirical risk and generalization error, respectively. We develop an understanding of data augmentation as regularization, which highlights the major features. As a result, data augmentation significantly reduces the generalization error, but meanwhile leads to a slightly higher empirical risk. On the assumption that data augmentation helps models converge to a better region, the model can benefit from a lower empirical risk achieved by a simple method, i.e., using less-augmented data to refine the model trained on fully-augmented data. Our approach achieves consistent accuracy gain on a few standard image classification benchmarks, and the gain transfers to object detection.