Point cloud is an important type of geometric data structure. Due to its
irregular format, most researchers transform such data to regular 3D voxel
grids or collections of images. This, however, renders data unnecessarily
voluminous and causes issues. In this paper, we design a novel type of neural
network that directly consumes point clouds and well respects the permutation
invariance of points in the input. Our network, named PointNet, provides a
unified architecture for applications ranging from object classification, part
segmentation, to scene semantic parsing. Though simple, PointNet is highly
efficient and effective. Empirically, it shows strong performance on par or
even better than state of the art. Theoretically, we provide analysis towards
understanding of what the network has learnt and why the network is robust with
respect to input perturbation and corruption.
Description
[1612.00593] PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation
%0 Generic
%1 qi2016pointnet
%A Qi, Charles R.
%A Su, Hao
%A Mo, Kaichun
%A Guibas, Leonidas J.
%D 2016
%K 2016 3D arxiv classification deep-learning paper point-cloud segmentation stanford
%T PointNet: Deep Learning on Point Sets for 3D Classification and
Segmentation
%U http://arxiv.org/abs/1612.00593
%X Point cloud is an important type of geometric data structure. Due to its
irregular format, most researchers transform such data to regular 3D voxel
grids or collections of images. This, however, renders data unnecessarily
voluminous and causes issues. In this paper, we design a novel type of neural
network that directly consumes point clouds and well respects the permutation
invariance of points in the input. Our network, named PointNet, provides a
unified architecture for applications ranging from object classification, part
segmentation, to scene semantic parsing. Though simple, PointNet is highly
efficient and effective. Empirically, it shows strong performance on par or
even better than state of the art. Theoretically, we provide analysis towards
understanding of what the network has learnt and why the network is robust with
respect to input perturbation and corruption.
@misc{qi2016pointnet,
abstract = {Point cloud is an important type of geometric data structure. Due to its
irregular format, most researchers transform such data to regular 3D voxel
grids or collections of images. This, however, renders data unnecessarily
voluminous and causes issues. In this paper, we design a novel type of neural
network that directly consumes point clouds and well respects the permutation
invariance of points in the input. Our network, named PointNet, provides a
unified architecture for applications ranging from object classification, part
segmentation, to scene semantic parsing. Though simple, PointNet is highly
efficient and effective. Empirically, it shows strong performance on par or
even better than state of the art. Theoretically, we provide analysis towards
understanding of what the network has learnt and why the network is robust with
respect to input perturbation and corruption.},
added-at = {2018-07-20T09:08:17.000+0200},
author = {Qi, Charles R. and Su, Hao and Mo, Kaichun and Guibas, Leonidas J.},
biburl = {https://www.bibsonomy.org/bibtex/2987e31c1772727d8edf002a9fd4b9164/analyst},
description = {[1612.00593] PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation},
interhash = {e2b76e0a980b5e6edb66a091d32ba0f5},
intrahash = {987e31c1772727d8edf002a9fd4b9164},
keywords = {2016 3D arxiv classification deep-learning paper point-cloud segmentation stanford},
note = {cite arxiv:1612.00593Comment: CVPR 2017},
timestamp = {2018-07-20T09:13:55.000+0200},
title = {PointNet: Deep Learning on Point Sets for 3D Classification and
Segmentation},
url = {http://arxiv.org/abs/1612.00593},
year = 2016
}