Vanishing Twin GAN: How training a weak Generative Adversarial Network can improve semi-supervised image classification

TL;DR

The paper introduces Vanishing Twin GAN, a novel training method that uses a weak GAN alongside a regular GAN to enhance semi-supervised image classification, especially when classes are similar.

Contribution

It proposes a new training approach that leverages a weak GAN to improve classification accuracy in semi-supervised learning with similar classes.

Findings

Improved classification accuracy in semi-supervised tasks.

Effective handling of similar or overlapping classes.

Enhanced generalization in GAN-based semi-supervised learning.

Abstract

Generative Adversarial Networks can learn the mapping of random noise to realistic images in a semi-supervised framework. This mapping ability can be used for semi-supervised image classification to detect images of an unknown class where there is no training data to be used for supervised classification. However, if the unknown class shares similar characteristics to the known class(es), GANs can learn to generalize and generate images that look like both classes. This generalization ability can hinder the classification performance. In this work, we propose the Vanishing Twin GAN. By training a weak GAN and using its generated output image parallel to the regular GAN, the Vanishing Twin training improves semi-supervised image classification where image similarity can hurt classification tasks.