P2P-Bridge: Diffusion Bridges for 3D Point Cloud Denoising

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Vogel, Mathias, Tateno, Keisuke, Pollefeys, Marc, Tombari, Federico, Rakotosaona, Marie-Julie, Engelmann, Francis
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916374445555712
author Vogel, Mathias
Tateno, Keisuke
Pollefeys, Marc
Tombari, Federico
Rakotosaona, Marie-Julie
Engelmann, Francis
author_facet Vogel, Mathias
Tateno, Keisuke
Pollefeys, Marc
Tombari, Federico
Rakotosaona, Marie-Julie
Engelmann, Francis
contents In this work, we tackle the task of point cloud denoising through a novel framework that adapts Diffusion Schrödinger bridges to points clouds. Unlike previous approaches that predict point-wise displacements from point features or learned noise distributions, our method learns an optimal transport plan between paired point clouds. Experiments on object datasets like PU-Net and real-world datasets such as ScanNet++ and ARKitScenes show that P2P-Bridge achieves significant improvements over existing methods. While our approach demonstrates strong results using only point coordinates, we also show that incorporating additional features, such as color information or point-wise DINOv2 features, further enhances the performance. Code and pretrained models are available at https://p2p-bridge.github.io.
format Preprint
id arxiv_https___arxiv_org_abs_2408_16325
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle P2P-Bridge: Diffusion Bridges for 3D Point Cloud Denoising
Vogel, Mathias
Tateno, Keisuke
Pollefeys, Marc
Tombari, Federico
Rakotosaona, Marie-Julie
Engelmann, Francis
Computer Vision and Pattern Recognition
In this work, we tackle the task of point cloud denoising through a novel framework that adapts Diffusion Schrödinger bridges to points clouds. Unlike previous approaches that predict point-wise displacements from point features or learned noise distributions, our method learns an optimal transport plan between paired point clouds. Experiments on object datasets like PU-Net and real-world datasets such as ScanNet++ and ARKitScenes show that P2P-Bridge achieves significant improvements over existing methods. While our approach demonstrates strong results using only point coordinates, we also show that incorporating additional features, such as color information or point-wise DINOv2 features, further enhances the performance. Code and pretrained models are available at https://p2p-bridge.github.io.
title P2P-Bridge: Diffusion Bridges for 3D Point Cloud Denoising
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2408.16325