Fast-converging Conditional Generative Adversarial Networks for Image Synthesis

TL;DR

This paper introduces FC-GAN, a conditional GAN that accelerates convergence in image synthesis by using an advanced auxiliary classifier to better distinguish real and fake data, leading to faster training without sacrificing image quality.

Contribution

The paper proposes a novel FC-GAN model with an enhanced auxiliary classifier that improves convergence speed in conditional GAN training.

Findings

Faster convergence rate compared to traditional GANs.

Competitive image quality in generated outputs.

Effective class differentiation during training.

Abstract

Building on top of the success of generative adversarial networks (GANs), conditional GANs attempt to better direct the data generation process by conditioning with certain additional information. Inspired by the most recent AC-GAN, in this paper we propose a fast-converging conditional GAN (FC-GAN). In addition to the real/fake classifier used in vanilla GANs, our discriminator has an advanced auxiliary classifier which distinguishes each real class from an extra `fake' class. The `fake' class avoids mixing generated data with real data, which can potentially confuse the classification of real data as AC-GAN does, and makes the advanced auxiliary classifier behave as another real/fake classifier. As a result, FC-GAN can accelerate the process of differentiation of all classes, thus boost the convergence speed. Experimental results on image synthesis demonstrate our model is competitive in the quality of images generated while achieving a faster convergence rate.