ReconFusion: 3D Reconstruction with Diffusion Priors

TL;DR

ReconFusion introduces a diffusion prior to enable high-quality 3D scene reconstruction from only a few images, significantly reducing capture time while maintaining realistic geometry and appearance.

Contribution

It presents a novel NeRF-based pipeline that uses diffusion priors trained on synthetic data to improve few-view 3D reconstruction quality.

Findings

Outperforms previous few-view NeRF methods in diverse real-world scenes

Synthesizes realistic geometry and textures in underconstrained regions

Effective across forward-facing and 360-degree scenes

Abstract

3D reconstruction methods such as Neural Radiance Fields (NeRFs) excel at rendering photorealistic novel views of complex scenes. However, recovering a high-quality NeRF typically requires tens to hundreds of input images, resulting in a time-consuming capture process. We present ReconFusion to reconstruct real-world scenes using only a few photos. Our approach leverages a diffusion prior for novel view synthesis, trained on synthetic and multiview datasets, which regularizes a NeRF-based 3D reconstruction pipeline at novel camera poses beyond those captured by the set of input images. Our method synthesizes realistic geometry and texture in underconstrained regions while preserving the appearance of observed regions. We perform an extensive evaluation across various real-world datasets, including forward-facing and 360-degree scenes, demonstrating significant performance improvements over previous few-view NeRF reconstruction approaches.