Surface normal estimation is a basic task for many point cloud processing algorithms. However, it can be challenging to capture the local geometry of the data, especially in presence of noise. Recently, deep learning approaches have shown promising results. Nevertheless, applying convolutional neural networks to point clouds is not straightforward, due to the irregular positioning of the points. In this paper, we propose a normal estimation method based on graph-convolutional neural networks to deal with such irregular point cloud domain. The graph-convolutional layers build hierarchies of localized features to solve the estimation problem. We show state-ofthe-art performance and robust results even in presence of noise.

Point Cloud Normal Estimation with Graph-Convolutional Neural Networks / Pistilli, Francesca; Fracastoro, Giulia; Valsesia, Diego; Magli, Enrico. - ELETTRONICO. - (2020), pp. 1-6. ((Intervento presentato al convegno 2020 IEEE International Conference on Multimedia & Expo - 3D Point Cloud Processing, Analysis, Compression, and Communication (PC-PACC) Workshop [10.1109/ICMEW46912.2020.9105972].

Point Cloud Normal Estimation with Graph-Convolutional Neural Networks

Pistilli, Francesca;Fracastoro, Giulia;Valsesia, Diego;Magli, Enrico
2020

Abstract

Surface normal estimation is a basic task for many point cloud processing algorithms. However, it can be challenging to capture the local geometry of the data, especially in presence of noise. Recently, deep learning approaches have shown promising results. Nevertheless, applying convolutional neural networks to point clouds is not straightforward, due to the irregular positioning of the points. In this paper, we propose a normal estimation method based on graph-convolutional neural networks to deal with such irregular point cloud domain. The graph-convolutional layers build hierarchies of localized features to solve the estimation problem. We show state-ofthe-art performance and robust results even in presence of noise.
978-1-7281-1485-9
File in questo prodotto:
File Dimensione Formato  
09105972.pdf

non disponibili

Tipologia: 2a Post-print versione editoriale / Version of Record
Licenza: Non Pubblico - Accesso privato/ristretto
Dimensione 1.38 MB
Formato Adobe PDF
1.38 MB Adobe PDF   Visualizza/Apri   Richiedi una copia
PC_normal.pdf

accesso aperto

Tipologia: 2. Post-print / Author's Accepted Manuscript
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 1.23 MB
Formato Adobe PDF
1.23 MB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

Caricamento pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: http://hdl.handle.net/11583/2844357