Unsupervised Image Generation with Infinite Generative Adversarial Networks

TL;DR

This paper introduces MIC-GANs, an unsupervised non-parametric approach that improves image generation by addressing common GAN issues like mode collapse and unstructured latent space, demonstrating superior performance across datasets.

Contribution

The paper proposes MIC-GANs, a novel mixture of infinite conditional GANs, which effectively structures the latent space and mitigates mode collapse without requiring extensive supervision.

Findings

MIC-GANs outperform state-of-the-art methods in image generation.

MIC-GANs effectively structure the latent space and avoid mode collapse.

The approach is adaptive, versatile, and robust across datasets.

Abstract

Image generation has been heavily investigated in computer vision, where one core research challenge is to generate images from arbitrarily complex distributions with little supervision. Generative Adversarial Networks (GANs) as an implicit approach have achieved great successes in this direction and therefore been employed widely. However, GANs are known to suffer from issues such as mode collapse, non-structured latent space, being unable to compute likelihoods, etc. In this paper, we propose a new unsupervised non-parametric method named mixture of infinite conditional GANs or MIC-GANs, to tackle several GAN issues together, aiming for image generation with parsimonious prior knowledge. Through comprehensive evaluations across different datasets, we show that MIC-GANs are effective in structuring the latent space and avoiding mode collapse, and outperform state-of-the-art methods. MICGANs are adaptive, versatile, and robust. They offer a promising solution to several well-known GAN issues. Code available: github.com/yinghdb/MICGANs.