[1]许煜濠,刘石坚,康朝明,等.三维深度学习网络的几何差异感知能力[J].福建理工大学学报,2023,21(06):592-597.[doi:10.3969/j.issn.1672-4348.2023.06.013]
　XU Yuhao,LIU Shijian,KANG Chaoming,et al.Geometric difference perception capabilities of 3D deep learning networks[J].Journal of Fujian University of Technology;,2023,21(06):592-597.[doi:10.3969/j.issn.1672-4348.2023.06.013]
点击复制

三维深度学习网络的几何差异感知能力()
分享到：

《福建理工大学学报》[ISSN:2097-3853/CN:35-1351/Z]

卷:
第21卷
期数:
2023年06期
页码:
592-597
栏目:
出版日期:
2023-12-25

文章信息/Info

Title:
Geometric difference perception capabilities of 3D deep learning networks
作者:
许煜濠; 刘石坚; 康朝明; 吴连杰; 邹峥
福建省大数据挖掘与应用技术重点实验室
Author(s):
XU Yuhao;  LIU Shijian;  KANG Chaoming;  WU Lianjie;  ZOU Zheng
Fujian Provincial Key Laboratory of Big Data Mining and Applications
关键词:
三角网格; 深度学习; 几何变换; 牙齿分割
Keywords:
triangular meshes; deep learning; geometric transformation; tooth segmentation
分类号:
TP301
DOI:
10.3969/j.issn.1672-4348.2023.06.013
文献标志码:
A
摘要:
在使用深度学习技术处理三角网格等三维数据时,如果网络不具备感知数据位置、朝向、尺寸等几何属性差异的能力,可能导致模型泛化能力不足、准确率偏低的后果。为解决该问题,在变换网络T?Net 的基础上,提出名为几何差异感知(geometric difference perception,GDP)的网络模块。其核心思想是通过多样化的样本训练,学习到一个变换矩阵,对高维特征进行规范化。通过以牙齿网格分割为任务的多项实验表明,GDP 能够有效应对三维数据的几何差异问题,避免其对模型性能造成的不良影响,对于网格分割等三维任务性能的提升具有重要意义。
Abstract:
When using deep learning techniques to process three-dimensional data like triangular meshes, if the network cannot perceive the differences of geometric properties, such as position, orientation and size of the data, the trained model may have poor generalization ability and low accuracy. To solve this problem, a network module called geometric difference perception (GDP) is proposed based on the transformation network named T-Net. The core idea of GDP is to learn a transformation matrix through training with diverse samples and use it to standardize high-dimensional features. Multiple experiments regarding tooth mesh segmentation tasks demonstrate that GDP can effectively address the geometric differences of three-dimensional data, and avoid its adverse effects on the performance of the model, which is of great significance for improving the performance of 3D tasks such as mesh segmentation.

参考文献/References:

［1］ CHARLES R Q，HAO S，MO K C，et al. PointNet：deep learning on point sets for 3D classification and segmentation［C］∥2017 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Honolulu，HI: IEEE，2017：652-660.［2］ WANG Y，SUN Y B，LIU Z W，et al. Dynamic graph CNN for learning on point clouds［J］. ACM Transactions on Graphics，38（5）: 1-12.［3］ ZHAO YE，ZHANG L M，LIU Y，et al. Two-stream graph convolutional network for intra-oral scanner image segmentation［J］. IEEE Transactions on Medical Imaging，2022，41（4）：826-835. ［4］ MA TA，YANG Y Z，ZHAI J C，et al. A tooth segmentation method based on multiple geometric feature learning［J］. Healthcare，2022，10（10）：2089. ［5］ QI C R，YI L，SU H，et al. PointNet++：deep hierarchical feature learning on point sets in a metric space［C］∥Proceedings of the 31st International Conference on Neural Information Processing Systems. New York：ACM，2017：5105-5114. ［6］ ZHAO Y，ZHANG L M，YANG C S，et al. 3D Dental model segmentation with graph attentional convolution network［J］. Pattern Recognition Letters，2021，152：79-85. ［7］ 张雅玲，于泽宽，何炳蔚. 基于GCNN的CBCT模拟口扫点云数据牙齿分割算法［J］. 计算机辅助设计与图形学学报，2020，32（7）：1162-1170. ［8］ BENHAMADOU A，SMAOUI O，REKIK A，et al. 3DTeethSeg22：3D teeth scan segmentation and labeling challenge［EB/OL］. 2023：arXiv：2305.18277.

相似文献/References:

[1]成晓元,凌静秀,黄继辉,等.土仓压力与掘进参数相关性分析及预测模型[J].福建理工大学学报,2022,20(01):13.[doi:10.3969/j.issn.1672-4348.2022.01.003]
　CHENG Xiaoyuan,LING Jingxiu,HUANG Jihui,et al.Correlation analysis and prediction model of chamber earth pressure and excavation parameters[J].Journal of Fujian University of Technology;,2022,20(06):13.[doi:10.3969/j.issn.1672-4348.2022.01.003]
[2]叶建华,唐辉,罗奋翔,等.基于改进MobileNet的咖啡豆缺陷检测[J].福建理工大学学报,2023,21(03):257.[doi:10.3969/j.issn.1672-4348.2023.03.009]
　YE Jianhua,TANG Hui,LUO Fenxiang,et al.Coffee bean defect detection based on improved MobileNet[J].Journal of Fujian University of Technology;,2023,21(06):257.[doi:10.3969/j.issn.1672-4348.2023.03.009]
[3]郭宝椿,李佐勇,陈健,等.融合长短时记忆与图结构学习的水库水位预测[J].福建理工大学学报,2024,22(01):90.[doi:10.3969/j.issn.2097-3853.2024.01.013]
　GUO Baochun,LI Zuoyong,CHEN Jian,et al.Reservoir level prediction via integrating long short-term memory and graph structure learning[J].Journal of Fujian University of Technology;,2024,22(06):90.[doi:10.3969/j.issn.2097-3853.2024.01.013]
[4]于明源,周景亮,曾绍锋,等.基于YOLOv8改进的轴承表面缺陷检测方法[J].福建理工大学学报,2024,22(03):280.[doi:10.3969/j.issn.2097-3853.2024.03.011]
　YU Mingyuan,ZHOU Jingliang,ZENG Shaofeng,et al.Improved bearing surface defect detection method based on YOLOv8[J].Journal of Fujian University of Technology;,2024,22(06):280.[doi:10.3969/j.issn.2097-3853.2024.03.011]

更新日期/Last Update: 2023-12-25