MGGAN: Solving Mode Collapse using Manifold Guided Training

TL;DR

MGGAN introduces a guidance network to prevent mode collapse in GANs, successfully capturing all data modes without sacrificing image quality, and is adaptable to various GAN architectures.

Contribution

The paper presents MGGAN, a novel method that uses a guidance network to address mode collapse in GANs while maintaining image quality.

Findings

Resolves mode collapse effectively.

Preserves high image quality.

Easily extendable to existing GANs.

Abstract

Mode collapse is a critical problem in training generative adversarial networks. To alleviate mode collapse, several recent studies introduce new objective functions, network architectures or alternative training schemes. However, their achievement is often the result of sacrificing the image quality. In this paper, we propose a new algorithm, namely a manifold guided generative adversarial network (MGGAN), which leverages a guidance network on existing GAN architecture to induce generator learning all modes of data distribution. Based on extensive evaluations, we show that our algorithm resolves mode collapse without losing image quality. In particular, we demonstrate that our algorithm is easily extendable to various existing GANs. Experimental analysis justifies that the proposed algorithm is an effective and efficient tool for training GANs.