MISS GAN: A Multi-IlluStrator Style Generative Adversarial Network for image to illustration translation

TL;DR

MISS GAN introduces a multi-style unsupervised image-to-illustration translation framework that efficiently handles diverse styles with a single trained generator, preserving content across various illustrators.

Contribution

It proposes a novel multi-illustrator style GAN that supports multiple styles in a single model, overcoming limitations of existing methods requiring multiple or image-specific networks.

Findings

Supports diverse styles with one generator

Preserves content while transferring styles

Handles multiple illustrators' styles effectively

Abstract

Unsupervised style transfer that supports diverse input styles using only one trained generator is a challenging and interesting task in computer vision. This paper proposes a Multi-IlluStrator Style Generative Adversarial Network (MISS GAN) that is a multi-style framework for unsupervised image-to-illustration translation, which can generate styled yet content preserving images. The illustrations dataset is a challenging one since it is comprised of illustrations of seven different illustrators, hence contains diverse styles. Existing methods require to train several generators (as the number of illustrators) to handle the different illustrators' styles, which limits their practical usage, or require to train an image specific network, which ignores the style information provided in other images of the illustrator. MISS GAN is both input image specific and uses the information of other images using only one trained model.