ELF-Nets: Deep learning on point clouds using extended laplacian filter

Seon Ho Lee, Han Ul Kim, Chang Su Kim

Research output: Contribution to journalArticle

Abstract

We propose a deep learning framework for various 3D vision tasks, which takes a point cloud as input. The convolution is a basic operator for feature extraction in deep learning. However, it is not directly applicable to a point cloud, which is an irregular, unordered point set. This makes deep learning on point clouds challenging. To address this issue, we propose the extended Laplacian filter (ELF) for point clouds, which adopts the design principles of discrete Laplacian filters in 2D image processing. In other words, ELF extends the Laplacian filters and has the following two properties: 1) it is a two-state filter using two filter matrices (one for a center point and the other for neighboring points), and 2) it employs a scalar weighting function to predict the relative importance of the neighboring points. Then, we develop ELF-Nets, which consist of ELF convolution layers and fully connected layers. Experimental results demonstrate that the proposed ELF-Nets are capable of recognizing the 3D shape of a point cloud effectively and efficiently. In particular, ELF-Nets provide better or comparable performances than the state-of-the-art techniques in both object classification and part segmentation tasks.

Original languageEnglish
Article number8884200
Pages (from-to)156569-156581
Number of pages13
JournalIEEE Access
Volume7
DOIs
Publication statusPublished - 2019 Jan 1

Fingerprint

Convolution
Mathematical operators
Feature extraction
Image processing
Deep learning

Keywords

  • 3D deep learning
  • convolutional neural network
  • Laplacian filter
  • object classification
  • Point cloud
  • semantic part segmentation

ASJC Scopus subject areas

  • Computer Science(all)
  • Materials Science(all)
  • Engineering(all)

Cite this

ELF-Nets : Deep learning on point clouds using extended laplacian filter. / Lee, Seon Ho; Kim, Han Ul; Kim, Chang Su.

In: IEEE Access, Vol. 7, 8884200, 01.01.2019, p. 156569-156581.

Research output: Contribution to journalArticle

Lee, Seon Ho ; Kim, Han Ul ; Kim, Chang Su. / ELF-Nets : Deep learning on point clouds using extended laplacian filter. In: IEEE Access. 2019 ; Vol. 7. pp. 156569-156581.
@article{6e0760fc6a7f4181a63be972ba3c2c6d,
title = "ELF-Nets: Deep learning on point clouds using extended laplacian filter",
abstract = "We propose a deep learning framework for various 3D vision tasks, which takes a point cloud as input. The convolution is a basic operator for feature extraction in deep learning. However, it is not directly applicable to a point cloud, which is an irregular, unordered point set. This makes deep learning on point clouds challenging. To address this issue, we propose the extended Laplacian filter (ELF) for point clouds, which adopts the design principles of discrete Laplacian filters in 2D image processing. In other words, ELF extends the Laplacian filters and has the following two properties: 1) it is a two-state filter using two filter matrices (one for a center point and the other for neighboring points), and 2) it employs a scalar weighting function to predict the relative importance of the neighboring points. Then, we develop ELF-Nets, which consist of ELF convolution layers and fully connected layers. Experimental results demonstrate that the proposed ELF-Nets are capable of recognizing the 3D shape of a point cloud effectively and efficiently. In particular, ELF-Nets provide better or comparable performances than the state-of-the-art techniques in both object classification and part segmentation tasks.",
keywords = "3D deep learning, convolutional neural network, Laplacian filter, object classification, Point cloud, semantic part segmentation",
author = "Lee, {Seon Ho} and Kim, {Han Ul} and Kim, {Chang Su}",
year = "2019",
month = "1",
day = "1",
doi = "10.1109/ACCESS.2019.2949785",
language = "English",
volume = "7",
pages = "156569--156581",
journal = "IEEE Access",
issn = "2169-3536",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - ELF-Nets

T2 - Deep learning on point clouds using extended laplacian filter

AU - Lee, Seon Ho

AU - Kim, Han Ul

AU - Kim, Chang Su

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We propose a deep learning framework for various 3D vision tasks, which takes a point cloud as input. The convolution is a basic operator for feature extraction in deep learning. However, it is not directly applicable to a point cloud, which is an irregular, unordered point set. This makes deep learning on point clouds challenging. To address this issue, we propose the extended Laplacian filter (ELF) for point clouds, which adopts the design principles of discrete Laplacian filters in 2D image processing. In other words, ELF extends the Laplacian filters and has the following two properties: 1) it is a two-state filter using two filter matrices (one for a center point and the other for neighboring points), and 2) it employs a scalar weighting function to predict the relative importance of the neighboring points. Then, we develop ELF-Nets, which consist of ELF convolution layers and fully connected layers. Experimental results demonstrate that the proposed ELF-Nets are capable of recognizing the 3D shape of a point cloud effectively and efficiently. In particular, ELF-Nets provide better or comparable performances than the state-of-the-art techniques in both object classification and part segmentation tasks.

AB - We propose a deep learning framework for various 3D vision tasks, which takes a point cloud as input. The convolution is a basic operator for feature extraction in deep learning. However, it is not directly applicable to a point cloud, which is an irregular, unordered point set. This makes deep learning on point clouds challenging. To address this issue, we propose the extended Laplacian filter (ELF) for point clouds, which adopts the design principles of discrete Laplacian filters in 2D image processing. In other words, ELF extends the Laplacian filters and has the following two properties: 1) it is a two-state filter using two filter matrices (one for a center point and the other for neighboring points), and 2) it employs a scalar weighting function to predict the relative importance of the neighboring points. Then, we develop ELF-Nets, which consist of ELF convolution layers and fully connected layers. Experimental results demonstrate that the proposed ELF-Nets are capable of recognizing the 3D shape of a point cloud effectively and efficiently. In particular, ELF-Nets provide better or comparable performances than the state-of-the-art techniques in both object classification and part segmentation tasks.

KW - 3D deep learning

KW - convolutional neural network

KW - Laplacian filter

KW - object classification

KW - Point cloud

KW - semantic part segmentation

UR - http://www.scopus.com/inward/record.url?scp=85074656400&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85074656400&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2019.2949785

DO - 10.1109/ACCESS.2019.2949785

M3 - Article

AN - SCOPUS:85074656400

VL - 7

SP - 156569

EP - 156581

JO - IEEE Access

JF - IEEE Access

SN - 2169-3536

M1 - 8884200

ER -