LocoGAN -- Locally Convolutional GAN

TL;DR

LocoGAN introduces a fully convolutional GAN that generates arbitrarily sized images by processing local sub-images, utilizing position channels to produce periodic and infinitely long images, advancing flexible image synthesis.

Contribution

It presents a novel fully convolutional GAN architecture with local processing and position channels, enabling flexible, high-resolution, and periodic image generation.

Findings

Can produce images of arbitrary dimensions.

Supports generation of periodic and panoramic images.

Uses local processing for scalable image synthesis.

Abstract

In the paper we construct a fully convolutional GAN model: LocoGAN, which latent space is given by noise-like images of possibly different resolutions. The learning is local, i.e. we process not the whole noise-like image, but the sub-images of a fixed size. As a consequence LocoGAN can produce images of arbitrary dimensions e.g. LSUN bedroom data set. Another advantage of our approach comes from the fact that we use the position channels, which allows the generation of fully periodic (e.g. cylindrical panoramic images) or almost periodic ,,infinitely long" images (e.g. wall-papers).