Multiple Style Transfer via Variational AutoEncoder

TL;DR

This paper introduces ST-VAE, a Variational AutoEncoder that enables efficient and flexible multiple style transfer by projecting styles into a linear latent space for interpolation, outperforming existing methods.

Contribution

The paper presents a novel VAE-based approach for multiple style transfer that allows linear interpolation of styles, improving speed and flexibility over prior methods.

Findings

ST-VAE outperforms existing style transfer methods in quality.

ST-VAE is faster and more flexible than previous approaches.

It successfully performs both single and multiple style transfer on COCO dataset.

Abstract

Modern works on style transfer focus on transferring style from a single image. Recently, some approaches study multiple style transfer; these, however, are either too slow or fail to mix multiple styles. We propose ST-VAE, a Variational AutoEncoder for latent space-based style transfer. It performs multiple style transfer by projecting nonlinear styles to a linear latent space, enabling to merge styles via linear interpolation before transferring the new style to the content image. To evaluate ST-VAE, we experiment on COCO for single and multiple style transfer. We also present a case study revealing that ST-VAE outperforms other methods while being faster, flexible, and setting a new path for multiple style transfer.