Guiding InfoGAN with Semi-Supervision

TL;DR

This paper introduces ss-InfoGAN, a semi-supervised GAN that effectively uses minimal labels to generate semantically meaningful images, improves sample quality, and accelerates training.

Contribution

It presents a novel semi-supervised architecture based on InfoGAN that learns meaningful representations with very limited labeled data and enhances training efficiency.

Findings

Achieves higher quality image synthesis with minimal labels

Speeds up training convergence using small labeled datasets

Maintains disentanglement of latent variables without labels

Abstract

In this paper we propose a new semi-supervised GAN architecture (ss-InfoGAN) for image synthesis that leverages information from few labels (as little as 0.22%, max. 10% of the dataset) to learn semantically meaningful and controllable data representations where latent variables correspond to label categories. The architecture builds on Information Maximizing Generative Adversarial Networks (InfoGAN) and is shown to learn both continuous and categorical codes and achieves higher quality of synthetic samples compared to fully unsupervised settings. Furthermore, we show that using small amounts of labeled data speeds-up training convergence. The architecture maintains the ability to disentangle latent variables for which no labels are available. Finally, we contribute an information-theoretic reasoning on how introducing semi-supervision increases mutual information between synthetic and real data.