PointInfinity: Resolution-Invariant Point Diffusion Models

Zixuan Huang, Justin Johnson, Shoubhik Debnath, James M. Rehg, Chao Yuan Wu

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

Abstract

We present PointInfinity, an efficient family of point cloud diffusion models. Our core idea is to use a transformer-based architecture with a fixed-size, resolution-invariant latent representation. This enables efficient training with low-resolution point clouds, while allowing high-resolution point clouds to be generated during inference. More importantly, we show that scaling the test-time resolution beyond the training resolution improves the fidelity of generated point clouds and surfaces. We analyze this phenomenon and draw a link to classifier-free guidance commonly used in diffusion models, demonstrating that both allow trading off fidelity and variability during inference. Experiments on CO3D show that PointInfinity can efficiently generate high-resolution point clouds (up to 131k points, 31× more than Point-E) with state-of-the-art quality.

Original languageEnglish (US)
Title of host publicationProceedings - 2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024
PublisherIEEE Computer Society
Pages10050-10060
Number of pages11
ISBN (Electronic)9798350353006
DOIs
StatePublished - 2024
Event2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024 - Seattle, United States
Duration: Jun 16 2024Jun 22 2024

Publication series

NameProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
ISSN (Print)1063-6919

Conference

Conference2024 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2024
Country/TerritoryUnited States
CitySeattle
Period6/16/246/22/24

Keywords

  • 3D Diffusion Model
  • 3D Generation
  • 3D reconstruction
  • 3D Vision
  • Computer Vision
  • Deep Learning

ASJC Scopus subject areas

  • Software
  • Computer Vision and Pattern Recognition

Fingerprint

Dive into the research topics of 'PointInfinity: Resolution-Invariant Point Diffusion Models'. Together they form a unique fingerprint.

Cite this