Detecting 3D Points of Interest Using Projective Neural Networks

被引:4
作者
Shu, Zhenyu [1 ,2 ]
Yang, Sipeng [3 ]
Xin, Shiqing [4 ]
Pang, Chaoyi [1 ,2 ]
Jin, Xiaogang [5 ]
Kavan, Ladislav [6 ]
Liu, Ligang [7 ]
机构
[1] NingboTech Univ, Sch Comp & Data Engn, Ningbo 315100, Peoples R China
[2] Zhejiang Univ, Ningbo Inst, Ningbo 315100, Peoples R China
[3] Zhejiang Univ, Sch Mech Engn, Hangzhou 310058, Peoples R China
[4] Shandong Univ, Sch Comp Sci & Technol, Jinan 250100, Shandong, Peoples R China
[5] Zhejiang Univ, State Key Lab CAD & CG, Hangzhou 310058, Peoples R China
[6] Univ Utah, Sch Comp, Salt Lake City, UT 84112 USA
[7] Univ Sci & Technol China, Sch Math Sci, Graph & Geometr Comp Lab, Hefei 230026, Anhui, Peoples R China
基金
中国国家自然科学基金; 美国国家科学基金会;
关键词
Shape; Three-dimensional displays; Feature extraction; Two dimensional displays; Neural networks; Training; Cameras; 3D shapes; Point of interest; Convolutional neural networks; SALIENCY DETECTION; MESH SALIENCY;
D O I
10.1109/TMM.2021.3070977
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Detecting points of interest on 3D shapes is a fundamental research problem in geometry processing. Due to the complicated relationship between points of interest and their geometric features, detecting points of interest on any given 3D shape remains challenging. Due to the lack of training data, previous data-driven methods for detecting 3D points of interest mainly focus on utilizing hand-crafted geometric features to predict the probabilities of each point being a POI, which greatly limits detection performance. In this paper, we propose a novel algorithm for detecting 3D points of interest by using projective neural networks. Our method first projects the labeled training 3D shapes into multiple 2D views and then learns the required features from the 2D views in an end-to-end fashion. The points of interest on test 3D shapes are then automatically detected by applying the learned neural network and our improved density peak clustering. Our method relies neither on hand-crafted feature descriptors nor a large quantity of expensive 3D training data to obtain satisfactory results. Experimental results show significantly superior detection performance of our method over the state-of-the-art methods.
引用
收藏
页码:1637 / 1650
页数:14
相关论文
共 50 条
[41]   Multi-Modal Segmentation of 3D Brain Scans Using Neural Networks [J].
Zopes, Jonathan ;
Platscher, Moritz ;
Paganucci, Silvio ;
Federau, Christian .
FRONTIERS IN NEUROLOGY, 2021, 12
[42]   Reconstruction of the Problem of 3D Electrical Impedance Tomography by using Artificial Neural Networks [J].
Duran, Guilherme C. ;
Sato, Andre K. ;
Ueda, Edson K. ;
Martins, Thiago C. ;
Takimoto, Rogerio Y. ;
Tsuzuki, Marcos S. G. .
2021 14TH IEEE INTERNATIONAL CONFERENCE ON INDUSTRY APPLICATIONS (INDUSCON), 2021, :994-998
[43]   Depth-image Coding Using Neural Networks for 3D Video Transmission [J].
Lin, Yih-Chuan ;
Hsu, Pu-Jian .
PIERS 2013 STOCKHOLM: PROGRESS IN ELECTROMAGNETICS RESEARCH SYMPOSIUM, 2013, :479-483
[44]   Automatic sequence of 3D point data for surface fitting using neural networks [J].
He Xueming ;
Li Chenggang ;
Hu Yujin ;
Zhang Rong ;
Yang, Simon X. ;
Mittal, Gauri S. .
COMPUTERS & INDUSTRIAL ENGINEERING, 2009, 57 (01) :408-418
[45]   MULTI-VIEW GAIT RECOGNITION USING 3D CONVOLUTIONAL NEURAL NETWORKS [J].
Wolf, Thomas ;
Babaee, Mohammadreza ;
Rigoll, Gerhard .
2016 IEEE INTERNATIONAL CONFERENCE ON IMAGE PROCESSING (ICIP), 2016, :4165-4169
[46]   3D Video-based Motion Capture using Convolutional Neural Networks [J].
ShangGuan, Huyuan ;
Mukundan, Ramakrishnan .
2017 INTERNATIONAL CONFERENCE ON IMAGE AND VISION COMPUTING NEW ZEALAND (IVCNZ), 2017,
[47]   Correspondence matching in unorganized 3D point clouds using Convolutional Neural Networks [J].
Pujol-Miro, Alba ;
Casas, Josep R. ;
Ruiz-Hidalgo, Javier .
IMAGE AND VISION COMPUTING, 2019, 83-84 :51-60
[48]   Exploratory Studies Detecting Secondary Structures in Medium Resolution 3D Cryo-EM Images Using Deep Convolutional Neural Networks [J].
Haslam, Devin ;
Zeng, Tao ;
Li, Rongjian ;
He, Jing .
ACM-BCB'18: PROCEEDINGS OF THE 2018 ACM INTERNATIONAL CONFERENCE ON BIOINFORMATICS, COMPUTATIONAL BIOLOGY, AND HEALTH INFORMATICS, 2018, :628-632
[49]   Adaptive 3D mesh reconstruction from dense unorganized weighted points using neural network [J].
Yan, LM ;
Yuan, YW ;
Zeng, XH .
PROCEEDINGS OF THE 2004 INTERNATIONAL CONFERENCE ON MACHINE LEARNING AND CYBERNETICS, VOLS 1-7, 2004, :3238-3242
[50]   Segmentation of 3D MRI Using 2D Convolutional Neural Networks in Infants' Brain [J].
Karimi, Hamed ;
Hamghalam, Mohammad .
MULTIMEDIA TOOLS AND APPLICATIONS, 2024, 83 (11) :33511-33526