GridNet: efficiently learning deep hierarchical representation for 3D point cloud understanding

被引：13

作者：

Wang, Huiqun ^{[1
]}

Huang, Di ^{[1
]}

Wang, Yunhong ^{[1
]}

机构：

[1] Beihang Univ, Sch Comp Sci & Engn, Lab Intelligent Recognit & Image Proc, Beijing 100191, Peoples R China

来源：

FRONTIERS OF COMPUTER SCIENCE | 2022年 / 16卷 / 01期

基金：

中国国家自然科学基金;

关键词：

3D point clouds; deep representations;

D O I：

10.1007/s11704-020-9521-2

中图分类号：

TP [自动化技术、计算机技术];

学科分类号：

0812 ;

摘要：

In this paper, we propose a novel and effective approach, namely GridNet, to hierarchically learn deep representation of 3D point clouds. It incorporates the ability of regular holistic description and fast data processing in a single framework, which is able to abstract powerful features progressively in an efficient way. Moreover, to capture more accurate internal geometry attributes, anchors are inferred within local neighborhoods, in contrast to the fixed or the sampled ones used in existing methods, and the learned features are thus more representative and discriminative to local point distribution. GridNet delivers very competitive results compared with the state of the art methods in both the object classification and segmentation tasks.

引用

页数：9