X-GAN: Improving Generative Adversarial Networks with ConveX Combinations

TitleX-GAN: Improving Generative Adversarial Networks with ConveX Combinations
Publication TypeConference Paper
Year of Publication2018
AuthorsBlum, O, Brattoli, B, Ommer, B
Conference NameOral at German Conference on Pattern Recognition (GCPR)
Conference LocationStuttgart, Germany
Keywordsdeep learning, generative adversarial network, generative model, variational auto-encoder
Abstract

Even though recent neural architectures for image generation are capable of producing photo-realistic results, the overall distributions of real and faked images still differ a lot. While the lack of a structured latent representation for GANs often results in mode collapse, VAEs enforce a prior to the latent space that leads to an unnatural representation of the underlying real distribution. We introduce a method that preserves the natural structure of the latent manifold. By utilizing neighboring relations within the set of discrete real samples, we reproduce the full continuous latent manifold. We propose a novel image generation network X-GAN that creates latent input vectors from random convex combinations of adjacent real samples. This way we ensure a structured and natural latent space by not requiring prior assumptions. In our experiments, we show that our model outperforms recent approaches in terms of the missing mode problem while maintaining a high image quality.

Citation Keyblum:GCPR:2018