Augmentation Inside the Network

TL;DR

This paper introduces a novel method called augmentation inside the network that applies data augmentation transformations to intermediate features, leading to faster and more accurate image classification.

Contribution

It proposes a new augmentation technique within the network that improves speed-accuracy trade-offs and enhances performance when combined with test-time augmentation.

Findings

Achieves similar accuracy to standard TTA with 30% faster speed on CIFAR-100.

Provides smoother speed-accuracy trade-offs.

Improves model performance when combined with TTA.

Abstract

In this paper, we present augmentation inside the network, a method that simulates data augmentation techniques for computer vision problems on intermediate features of a convolutional neural network. We perform these transformations, changing the data flow through the network, and sharing common computations when it is possible. Our method allows us to obtain smoother speed-accuracy trade-off adjustment and achieves better results than using standard test-time augmentation (TTA) techniques. Additionally, our approach can improve model performance even further when coupled with test-time augmentation. We validate our method on the ImageNet-2012 and CIFAR-100 datasets for image classification. We propose a modification that is 30% faster than the flip test-time augmentation and achieves the same results for CIFAR-100.