Resisting Large Data Variations via Introspective Transformation Network

TL;DR

This paper introduces a learnable transformation module within an introspective CNN to enhance robustness against large data variations, outperforming traditional data augmentation methods on multiple benchmark datasets.

Contribution

It presents a novel, principled method for training deep networks with improved resistance to data variations by embedding a learnable transformation in an introspective network.

Findings

Significant accuracy improvements on MNIST, affNIST, SVHN, CIFAR-10, and miniImageNet.

Enhanced robustness to large variations between training and testing data.

Outperforms standard data augmentation techniques.

Abstract

Training deep networks that generalize to a wide range of variations in test data is essential to building accurate and robust image classifiers. One standard strategy is to apply data augmentation to synthetically enlarge the training set. However, data augmentation is essentially a brute-force method which generates uniform samples from some pre-defined set of transformations. In this paper, we propose a principled approach to train networks with significantly improved resistance to large variations between training and testing data. This is achieved by embedding a learnable transformation module into the introspective network, which is a convolutional neural network (CNN) classifier empowered with generative capabilities. Our approach alternates between synthesizing pseudo-negative samples and transformed positive examples based on the current model, and optimizing model predictions on these synthesized samples. Experimental results verify that our approach significantly improves the ability of deep networks to resist large variations between training and testing data and achieves classification accuracy improvements on several benchmark datasets, including MNIST, affNIST, SVHN, CIFAR-10 and miniImageNet.