Rodin: A Generative Model for Sculpting 3D Digital Avatars Using Diffusion

TL;DR

This paper introduces Rodin, a novel diffusion-based 3D avatar generation model that efficiently produces high-quality, detailed, and editable neural radiance field avatars from text or images, overcoming computational challenges.

Contribution

Rodin employs a 2D feature map roll-out approach with 3D-aware diffusion and hierarchical synthesis, enabling efficient, high-fidelity 3D avatar generation and editing.

Findings

Generates highly detailed 3D avatars with realistic features.

Outperforms existing methods in quality and efficiency.

Supports text and image-based avatar creation and editing.

Abstract

This paper presents a 3D generative model that uses diffusion models to automatically generate 3D digital avatars represented as neural radiance fields. A significant challenge in generating such avatars is that the memory and processing costs in 3D are prohibitive for producing the rich details required for high-quality avatars. To tackle this problem we propose the roll-out diffusion network (Rodin), which represents a neural radiance field as multiple 2D feature maps and rolls out these maps into a single 2D feature plane within which we perform 3D-aware diffusion. The Rodin model brings the much-needed computational efficiency while preserving the integrity of diffusion in 3D by using 3D-aware convolution that attends to projected features in the 2D feature plane according to their original relationship in 3D. We also use latent conditioning to orchestrate the feature generation for global coherence, leading to high-fidelity avatars and enabling their semantic editing based on text prompts. Finally, we use hierarchical synthesis to further enhance details. The 3D avatars generated by our model compare favorably with those produced by existing generative techniques. We can generate highly detailed avatars with realistic hairstyles and facial hair like beards. We also demonstrate 3D avatar generation from image or text as well as text-guided editability.