HumanRef: Single Image to 3D Human Generation via Reference-Guided Diffusion
This addresses the problem of 3D human generation for applications like virtual reality or animation, but it is incremental as it builds on existing diffusion-based methods with specific improvements.
The paper tackles the problem of generating 3D human models from a single image, which is challenging due to inferring unseen textures and geometries while maintaining consistency, and proposes HumanRef with a novel reference-guided diffusion method that outperforms state-of-the-art methods in producing fine geometry, photorealistic textures, and view-consistent appearances.
Generating a 3D human model from a single reference image is challenging because it requires inferring textures and geometries in invisible views while maintaining consistency with the reference image. Previous methods utilizing 3D generative models are limited by the availability of 3D training data. Optimization-based methods that lift text-to-image diffusion models to 3D generation often fail to preserve the texture details of the reference image, resulting in inconsistent appearances in different views. In this paper, we propose HumanRef, a 3D human generation framework from a single-view input. To ensure the generated 3D model is photorealistic and consistent with the input image, HumanRef introduces a novel method called reference-guided score distillation sampling (Ref-SDS), which effectively incorporates image guidance into the generation process. Furthermore, we introduce region-aware attention to Ref-SDS, ensuring accurate correspondence between different body regions. Experimental results demonstrate that HumanRef outperforms state-of-the-art methods in generating 3D clothed humans with fine geometry, photorealistic textures, and view-consistent appearances.