Single-Stage Diffusion NeRF: A Unified Approach to 3D Generation and Reconstruction

ICCV 2023

1Tongji University   2Apple   3ETH Zürich   4UCSD   5University of Pennsylvania
*Work done during a remote internship with UCSD

TLDR: Trains 3D diffusion and NeRF jointly in a single stage
Comparable to/better than SoTAs on 3D generation and reconstruction


3D-aware image synthesis encompasses a variety of tasks, such as scene generation and novel view synthesis from images. Despite numerous task-specific methods, developing a comprehensive model remains challenging. In this paper, we present SSDNeRF, a unified approach that employs an expressive diffusion model to learn a generalizable prior of neural radiance fields (NeRF) from multi-view images of diverse objects. Previous studies have used two-stage approaches that rely on pretrained NeRFs as real data to train diffusion models. In contrast, we propose a new single-stage training paradigm with an end-to-end objective that jointly optimizes a NeRF auto-decoder and a latent diffusion model, enabling simultaneous 3D reconstruction and prior learning, even from sparsely available views. At test time, we can directly sample the diffusion prior for unconditional generation, or combine it with arbitrary observations of unseen objects for NeRF reconstruction. SSDNeRF demonstrates robust results comparable to or better than leading task-specific methods in unconditional generation and single/sparse-view 3D reconstruction.

Unconditional Generation

Trained on ABO Tables

Trained on ShapeNet-SRN Cars

Single-View Reconstruction

Inputs from ShapeNet-SRN Cars test set

Inputs from ShapeNet-SRN Chairs test set

[Sim-to-real] Inputs from KITTI real images (trained on SRN Cars)

Sparse-to-Dense Reconstruction

Failed to load image

Novel view synthesis quality (LPIPS) vs. number of input views, evaluated on SRN Cars.


    title={Single-Stage Diffusion NeRF: A Unified Approach to 3D Generation and Reconstruction}, 
    author={Hansheng Chen and Jiatao Gu and Anpei Chen and Wei Tian and Zhuowen Tu and Lingjie Liu and Hao Su},