Domain Enhanced Arbitrary Image Style Transfer via Contrastive Learning

TL;DR

This paper introduces CAST, a contrastive learning-based method for arbitrary image style transfer that learns style representations directly from image features, resulting in improved stylization quality and consistency.

Contribution

It proposes a novel contrastive learning framework for style representation in image transfer, surpassing second-order statistics methods in quality and artifact reduction.

Findings

Significantly better stylization results than state-of-the-art methods.

Effective learning of style distribution through contrastive learning.

Improved style consistency and reduced artifacts.

Abstract

In this work, we tackle the challenging problem of arbitrary image style transfer using a novel style feature representation learning method. A suitable style representation, as a key component in image stylization tasks, is essential to achieve satisfactory results. Existing deep neural network based approaches achieve reasonable results with the guidance from second-order statistics such as Gram matrix of content features. However, they do not leverage sufficient style information, which results in artifacts such as local distortions and style inconsistency. To address these issues, we propose to learn style representation directly from image features instead of their second-order statistics, by analyzing the similarities and differences between multiple styles and considering the style distribution. Specifically, we present Contrastive Arbitrary Style Transfer (CAST), which is a new style representation learning and style transfer method via contrastive learning. Our framework consists of three key components, i.e., a multi-layer style projector for style code encoding, a domain enhancement module for effective learning of style distribution, and a generative network for image style transfer. We conduct qualitative and quantitative evaluations comprehensively to demonstrate that our approach achieves significantly better results compared to those obtained via state-of-the-art methods. Code and models are available at https://github.com/zyxElsa/CAST_pytorch