What question did this study set out to answer?

The aim is to evaluate various data augmentation techniques for enhancing image classification performance using deep learning.

December 13, 2017Open Access

The Effectiveness of Data Augmentation in Image Classification using Deep Learning

Key Points

The aim is to evaluate various data augmentation techniques for enhancing image classification performance using deep learning.
Explored conventional techniques like cropping, rotating, and flipping on a limited ImageNet subset.
Compared traditional augmentation strategies with GAN-generated images for style diversity.
Developed a neural augmentation method that learns the best transformations to enhance classifiers.
Traditional transformations such as cropping and rotating showed significant effectiveness in boosting model performance.
GANs produced diverse image styles but varied in success across datasets.
Neural augmentation demonstrated potential improvements but also had notable limitations in specific scenarios.

Abstract

In this paper, we explore and compare multiple solutions to the problem of data augmentation in image classification. Previous work has demonstrated the effectiveness of data augmentation through simple techniques, such as cropping, rotating, and flipping input images. We artificially constrain our access to data to a small subset of the ImageNet dataset, and compare each data augmentation technique in turn. One of the more successful data augmentations strategies is the traditional transformations mentioned above. We also experiment with GANs to generate images of different styles. Finally, we propose a method to allow a neural net to learn augmentations that best improve the classifier, which we call neural augmentation. We discuss the successes and shortcomings of this method on various datasets.

Mark Helpful

Bookmark

Relay

View Full Paper