ArtFlow: Unbiased Image Style Transfer via Reversible Neural Flows

TL;DR

ArtFlow introduces a reversible neural flow-based method for universal style transfer that prevents content leak and maintains high-quality stylization through a lossless, unbiased feature transfer process.

Contribution

It proposes a novel reversible neural flow framework with an unbiased feature transfer module to address content leak in style transfer.

Findings

Achieves comparable style transfer quality to state-of-the-art methods.

Effectively prevents content leak during multiple stylization rounds.

Supports both forward and backward inference for lossless feature mapping.

Abstract

Universal style transfer retains styles from reference images in content images. While existing methods have achieved state-of-the-art style transfer performance, they are not aware of the content leak phenomenon that the image content may corrupt after several rounds of stylization process. In this paper, we propose ArtFlow to prevent content leak during universal style transfer. ArtFlow consists of reversible neural flows and an unbiased feature transfer module. It supports both forward and backward inferences and operates in a projection-transfer-reversion scheme. The forward inference projects input images into deep features, while the backward inference remaps deep features back to input images in a lossless and unbiased way. Extensive experiments demonstrate that ArtFlow achieves comparable performance to state-of-the-art style transfer methods while avoiding content leak.