Cosmos Propagation Network: Deep learning model for point cloud completion

Lin, Fangzhou1; Xu, Yajun2; Zhang, Ziming3; Gao, Chenyang4; Yamada, Kazunori D.1

2022-10-01

10.1016/j.neucom.2022.08.007

NEUROCOMPUTING   影响因子和分区

0925-2312

1872-8286

Point clouds measured by 3D scanning devices often have partially missing data due to the view positioning of the scanner. The missing data can reduce the performance of a point cloud in downstream tasks such as segmentation, location, and pose estimation. Consequently, 3D point cloud completion aims to predict the missing regions of incomplete objects for these fundamental 3D vision tasks. However, predicting the complete object can easily diminish the detail or structure of a measured region, which usually does not require repair. This study proposes a novel neural network architecture, Cosmos Propagation Network (CP-Net), for 3D point cloud completion. CP-Net extracts latent features in different scales from incomplete point clouds used as input. For point cloud generation, we propose a novel point expand method using a Mirror Expand module. Compared with existing methods, our Mirror Expand module introduces less information redundancy, which makes the distribution of points more reliable. CP-Net predicts the details of missing regions and maintains a clear general structure. The performance of CP-Net on several benchmarks was compared to that of current baseline methods. Compared to the existing methods, CP-Net showed the best performance for various metrics. Thus, CP-Net is expected to help address various problems related to 3D point cloud completion. Its source code is available at https://github.com/ark1234/CP-Net.(c) 2022 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

Shape completion 3D Point cloud Deep learning Neural networks Point propagation

SCI ; EI

英语

其他

NSF[CCF-2006738]

Computer Science

Computer Science, Artificial Intelligence

WOS:000843489800004

ELSEVIER

20223512647809

Backpropagation ; Benchmarking ; Deep neural networks ; Network architecture

Ergonomics and Human Factors Engineering:461.4 ; Artificial Intelligence:723.4 ; Optical Devices and Systems:741.3

COMPUTER SCIENCE

Web of Science

1.Tohoku Univ, Grad Sch Informat Sci, Dept Appl Informat Sci, Sendai, Miyagi 9808579, Japan
2.Hokkaido Univ, Fac Informat Sci & Technol, Dept Syst Sci & Informat, Sapporo, Hokkaido 0600814, Japan
3.Worcester Polytech Inst, ECE Dept & Data Sci & Robot Engn, Worcester, MA 01609 USA
4.Southern Univ Sci & Technol, Dept Comp Sci & Engn, Shenzhen, Peoples R China

Lin, Fangzhou,Xu, Yajun,Zhang, Ziming,et al. Cosmos Propagation Network: Deep learning model for point cloud completion[J]. NEUROCOMPUTING,2022,507:221-234.

Lin, Fangzhou,Xu, Yajun,Zhang, Ziming,Gao, Chenyang,&Yamada, Kazunori D..(2022).Cosmos Propagation Network: Deep learning model for point cloud completion.NEUROCOMPUTING,507,221-234.

Lin, Fangzhou,et al."Cosmos Propagation Network: Deep learning model for point cloud completion".NEUROCOMPUTING 507(2022):221-234.