MagicMan: Generative Novel View Synthesis of Humans with 3D-Aware Diffusion and Iterative Refinement

TL;DR

MagicMan is a novel human view synthesis method that combines 3D-aware diffusion, multi-view attention, and iterative refinement to produce consistent, high-quality multi-view images from a single reference image, improving 3D reconstruction.

Contribution

It introduces a multi-view diffusion model with hybrid attention, geometry-aware dual branches, and iterative refinement for better 3D human synthesis from a single image.

Findings

Outperforms existing methods in novel view synthesis.

Enhances 3D human reconstruction accuracy.

Achieves high consistency across generated views.

Abstract

Existing works in single-image human reconstruction suffer from weak generalizability due to insufficient training data or 3D inconsistencies for a lack of comprehensive multi-view knowledge. In this paper, we introduce MagicMan, a human-specific multi-view diffusion model designed to generate high-quality novel view images from a single reference image. As its core, we leverage a pre-trained 2D diffusion model as the generative prior for generalizability, with the parametric SMPL-X model as the 3D body prior to promote 3D awareness. To tackle the critical challenge of maintaining consistency while achieving dense multi-view generation for improved 3D human reconstruction, we first introduce hybrid multi-view attention to facilitate both efficient and thorough information interchange across different views. Additionally, we present a geometry-aware dual branch to perform concurrent generation in both RGB and normal domains, further enhancing consistency via geometry cues. Last but not least, to address ill-shaped issues arising from inaccurate SMPL-X estimation that conflicts with the reference image, we propose a novel iterative refinement strategy, which progressively optimizes SMPL-X accuracy while enhancing the quality and consistency of the generated multi-views. Extensive experimental results demonstrate that our method significantly outperforms existing approaches in both novel view synthesis and subsequent 3D human reconstruction tasks.