RenderDiffusion: Image Diffusion for 3D Reconstruction, Inpainting and Generation

Titas Anciukevicius, Zexiang Xu, Matthew Fisher, Paul Henderson, Hakan Bilen, Niloy J. Mitra, Paul Guerrero

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract / Description of output

Diffusion models currently achieve state-of-the-art performance for both conditional and unconditional image generation. However, so far, image diffusion models do not support tasks required for 3D understanding, such as view-consistent 3D generation or single-view object reconstruction. In this paper, we present RenderDiffusion, the first diffusion model for 3D generation and inference, trained using only monocular 2D supervision. Central to our method is a novel image denoising architecture that generates and renders an intermediate three-dimensional representation of a scene in each denoising step. This enforces a strong inductive structure within the diffusion process, providing a 3D consistent representation while only requiring 2D supervision. The resulting 3D representation can be rendered from any view. We evaluate RenderDiffusion on FFHQ, AFHQ, ShapeNet and CLEVR datasets, showing competitive performance for generation of 3D scenes and inference of 3D scenes from 2D images. Additionally, our diffusion-based approach allows us to use 2D inpainting to edit 3D scenes.
Original languageEnglish
Title of host publicationComputer Vision and Pattern Recognition 2023
Number of pages11
ISBN (Electronic)979-8-3503-0129-8
ISBN (Print)979-8-3503-0130-4
Publication statusPublished - 22 Aug 2023
EventThe IEEE/CVF Conference on Computer Vision and Pattern Recognition 2023 - Vancouver Convention Center, Vancouver, Canada
Duration: 18 Jun 202322 Jun 2023

Publication series

ISSN (Print)1063-6919
ISSN (Electronic)2575-7075


ConferenceThe IEEE/CVF Conference on Computer Vision and Pattern Recognition 2023
Abbreviated titleCVPR 2023
Internet address

Keywords / Materials (for Non-textual outputs)

  • cs.CV
  • cs.LG


Dive into the research topics of 'RenderDiffusion: Image Diffusion for 3D Reconstruction, Inpainting and Generation'. Together they form a unique fingerprint.

Cite this