Mixture of Style Experts for Diverse Image Stylization

TL;DR

StyleExpert is a semantic-aware diffusion stylization framework using a Mixture of Experts to handle diverse styles and preserve details, outperforming existing methods.

Contribution

Introduces StyleExpert, a novel MoE-based framework with a unified style encoder for diverse, semantic-aware image stylization.

Findings

Outperforms existing methods in preserving semantics and material details.

Generalizes well to unseen styles.

Handles styles across multiple semantic levels.

Abstract

Diffusion-based stylization has advanced significantly, yet existing methods are limited to color-driven transformations, neglecting complex semantics and material details. We introduce StyleExpert, a semantic-aware framework based on the Mixture of Experts (MoE). Our framework employs a unified style encoder, trained on our large-scale dataset of content-style-stylized triplets, to embed diverse styles into a consistent latent space. This embedding is then used to condition a similarity-aware gating mechanism, which dynamically routes styles to specialized experts within the MoE architecture. Leveraging this MoE architecture, our method adeptly handles diverse styles spanning multiple semantic levels, from shallow textures to deep semantics. Extensive experiments show that StyleExpert outperforms existing approaches in preserving semantics and material details, while generalizing to unseen styles. Our code and collected images are available at the project page: https://hh-lg.github.io/StyleExpert-Page/.