DiffStyle360: Diffusion-Based 360° Head Stylization via Style Fusion Attention

2025-11-27

AI Summary

Existing 3D head stylization methods rely on expensive per-instance optimization or single-style fine-tuning, suffering from poor generalization. This paper proposes the first single-image-guided 3D head stylization framework that ensures multi-view consistency and identity preservation. Our approach introduces three key innovations: (1) a latent-space style-fusion attention mechanism that adaptively balances structural fidelity and stylistic transfer; (2) an appearance-structure disentanglement module enabling cross-domain style generalization; and (3) a 3D-aware diffusion architecture built upon DiffPortrait360, enhanced by multi-view data synthesized via 3D GANs and a temperature-controlled key-scaling strategy for precise style intensity control. Experiments on FFHQ and RenderMe360 demonstrate significant improvements over state-of-the-art GAN- and diffusion-based methods across diverse artistic styles, achieving superior style fidelity and multi-view visual consistency.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Generative Models & Autoencoders
• Natural Language Processing: Sentiment Analysis, Stylistic Analysis, and Argument Mining

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

3D head stylization has emerged as a key technique for reimagining realistic human heads in various artistic forms, enabling expressive character design and creative visual experiences in digital media. Despite the progress in 3D-aware generation, existing 3D head stylization methods often rely on computationally expensive optimization or domain-specific fine-tuning to adapt to new styles. To address these limitations, we propose DiffStyle360, a diffusion-based framework capable of producing multi-view consistent, identity-preserving 3D head stylizations across diverse artistic domains given a single style reference image, without requiring per-style training. Building upon the 3D-aware DiffPortrait360 architecture, our approach introduces two key components: the Style Appearance Module, which disentangles style from content, and the Style Fusion Attention mechanism, which adaptively balances structure preservation and stylization fidelity in the latent space. Furthermore, we employ a 3D GAN-generated multi-view dataset for robust fine-tuning and introduce a temperaturebased key scaling strategy to control stylization intensity during inference. Extensive experiments on FFHQ and RenderMe360 demonstrate that DiffStyle360 achieves superior style quality, outperforming state-of-the-art GAN- and diffusion-based stylization methods across challenging style domains.

Problem

• Generates 3D head stylizations from single style images
• Eliminates per-style training for diverse artistic domains
• Ensures multi-view consistency and identity preservation

Innovation

• Diffusion-based framework for 3D head stylization
• Style Fusion Attention balances structure and stylization
• No per-style training needed with single reference image