广东工业大学学报 ›› 2023, Vol. 40 ›› Issue (04): 18-23.doi: 10.12052/gdutxb.220145

• 计算机科学与技术 • 上一篇    下一篇

基于通道分离机制的双分支点云处理网络

钟耿君, 李东   

  1. 广东工业大学 自动化学院，广东 广州 510006
  • 收稿日期:2022-09-19 出版日期:2023-07-25 发布日期:2023-08-02
  • 通信作者: 李东(1983–)， 男，副教授，博士，硕士生导师，主要研究方向为模式识别、机器学习、人脸识别和机器视觉，E-mail：dong.li@gdut.edu.cn
  • 作者简介:钟耿君(1997–)， 男，硕士研究生，主要研究方向为点云处理、模式识别
  • 基金资助:
    广东省自然科学基金资助项目(2021A1515011867)

A Channel-splited Based Dual-branch Block for 3D Point Cloud Processing

Zhong Geng-jun, Li Dong   

  1. School of Automation, Guangdong University of Technology, Guangzhou 510006, China
  • Received:2022-09-19 Online:2023-07-25 Published:2023-08-02

HTML

PDF (PC)

722

摘要： 本文在PointMLP方法的基础上，提出了基于通道分离机制的双分支网络模块(Channel-splited Based Dual-branch Block，CDBlock) 。CDBlock将输入特征在通道维度上切分成两组特征，并将它们输入到双分支网络模块中的不同网络分支上。具体地，双分支网络模块包括轻量网络分支和深层网络分支。轻量网络分支由残差多层感知机 (Multi-layer Perceptron, MLP) 结构组成，负责提取浅层特征信息。深层网络分支由瓶颈网络结构组成，负责提取深层语义信息。CDBlock的引入提升了网络的对点云数据的特征提取能力和学习能力，有效地提高了模型的鲁棒性。本文方法在点云分类数据集ScanObjectNN上进行了验证，总体精度和类别平均精度分别达到了86.2%和84.97%，优于PointMLP。此外，本文方法在点云分割数据集ShapeNetPart上也取得了具有竞争力的结果。相比于PointMLP，本文方法在使用更少的参数量和计算量情况下取得了更优异的结果。

关键词: 三维点云, 深度学习, 多分支, 残差网络

Abstract: In this paper, we propose a channel-split-based dual-branch block (CDBlock) based on the PointMLP method, which splits input features into two groups along the channel dimension and feeds them into different branches of the dual-branch network module. Specifically, our proposed CDBlock consists of a lightweight branch and a heavyweight branch. The lightweight branch, which uses the structure of the residual MLP, is designed to extract coarse semantic features. The heavyweight branch, which uses the bottleneck network as backbone, is responsible for extracting more deeper and distinguishable information. By doing this, our proposed CDBlock effectively improves the network's feature extraction and learning capabilities. Experimental results show that our proposed method outperforms the existing PointMLP by achieving an overall accuracy of 86.2% and a class average accuracy of 84.97% on ScanObjectNN dataset. In particular, our approach achieves better results with using fewer parameters and computational cost than the PointMLP. Additionally, our approach also achieves encouraging performance on the ShapeNetPart dataset.

Key words: 3D point cloud, deep learning, multi-branch, residual network

中图分类号: 

  • TP391.4
[1] QI C R, SU H, MO K, et al. Pointnet: deep learning on point sets for 3D classification and segmentation[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI: IEEE, 2017: 652-660.
[2] QI C R, YI L, SU H, et al. Pointnet++: deep hierarchical feature learning on point sets in a metric space[J]. Advances in Neural Information Processing Systems, 2017, 30: 5099-5108.
[3] THOMAS H, QI C R, DESCHAUD J E, et al. Kpconv: flexible and deformable convolution for point clouds[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Long Beach: IEEE, 2019: 6411-6420.
[4] WU W, QI Z, FUXIN L. Pointconv: deep convolutional networks on 3D point clouds[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Seoul: IEEE, 2019: 9621-9630.
[5] XU M, DING R, ZHAO H, et al. Paconv: position adaptive convolution with dynamic kernel assembling on point clouds[C]//Proceedings of the IEEE/ CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2021: 3173-3182.
[6] RAN H, LIU J, WANG C. Surface representation for point clouds[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 18942-18952.
[7] LIU Z, HU H, CAO Y, et al. A closer look at local aggregation operators in point cloud analysis[C]// European Conference on Computer Vision. Glasgow: ECCV, 2020: 326-342.
[8] DOSOVITSKIY A, BEYER L, KOLESNIKOV A, et al. An image is worth 16×16 words: transformers for image recognition at scale[C]//International Conference on Learning Representations. Cambridge: ICLR, 2021: 1-6.
[9] ZHAO H, JIANG L, JIA J, et al. Point transformer[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Montreal: IEEE, 2021: 16259-16268.
[10] GUO M H, CAI J X, LIU Z N, et al. Pct: point cloud transformer[J]. Computational Visual Media, 2021, 7(2): 187-199.
[11] XU MA, CAN Q, HAOXUAN Y, et al. Rethinking network design and local geometry in point cloud: a simple residual MLP framework[C]//International Conference on Learning Representations. Kigali: ICLR, 2022: 1-9.
[12] ZHANG H, WU C, ZHANG Z, et al. Resnest: split-attention networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. New Orleans: IEEE, 2022: 2736-2746.
[13] SZEGEDY C, LIU W, JIA Y, et al. Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Boston: IEEE, 2015: 1-9.
[14] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[15] UY M A, PHAM Q H, HUA B S, et al. Revisiting point cloud classification: a new benchmark dataset and classification model on real-world data[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. Seoul: IEEE, 2019: 1588-1597.
[16] YI L, KIM V G, CEYLAN D, et al. A scalable active framework for region annotation in 3D shape collections[J]. ACM Transactions on Graphics(ToG), 2016, 35(6): 1.
[17] LI Y, BU R, SUN M, et al. Pointcnn: convolution on x-transformed points[J]. Advances in Neural Information Processing Systems, 2018, 31: 1-11.
[1] 温雯, 刘莹, 蔡瑞初, 郝志峰. 面向多粒度交通流预测的时空深度回归模型[J]. 广东工业大学学报, 2023, 40(04): 1-8.
[2] 金宇凯, 李志生, 欧耀春, 张华刚, 曾江毅, 陈搏超. 基于多阶段聚类的PM2.5质量浓度预测及对比研究[J]. 广东工业大学学报, 2023, 40(03): 17-24.
[3] 吴俊贤, 何元烈. 基于通道注意力的自监督深度估计方法[J]. 广东工业大学学报, 2023, 40(02): 22-29.
[4] 刘冬宁, 王子奇, 曾艳姣, 文福燕, 王洋. 基于复合编码特征LSTM的基因甲基化位点预测方法[J]. 广东工业大学学报, 2023, 40(01): 1-9.
[5] 徐伟锋, 蔡述庭, 熊晓明. 基于深度特征的单目视觉惯导里程计[J]. 广东工业大学学报, 2023, 40(01): 56-60,76.
[6] 刘洪伟, 林伟振, 温展明, 陈燕君, 易闽琦. 基于MABM的消费者情感倾向识别模型——以电影评论为例[J]. 广东工业大学学报, 2022, 39(06): 1-9.
[7] 章云, 王晓东. 基于受限样本的深度学习综述与思考[J]. 广东工业大学学报, 2022, 39(05): 1-8.
[8] 郑佳碧, 杨振国, 刘文印. 基于细粒度混杂平衡的营销效果评估方法[J]. 广东工业大学学报, 2022, 39(02): 55-61.
[9] 杨积升, 章云, 李东. 点云目标检测残差投票网络[J]. 广东工业大学学报, 2022, 39(01): 56-62.
[10] Gary Yen, 栗波, 谢胜利. 地球流体动力学模型恢复的长短期记忆网络渐进优化方法[J]. 广东工业大学学报, 2021, 38(06): 1-8.
[11] 赖峻, 刘震宇, 刘圣海. 基于全局数据混洗的小样本数据预测方法[J]. 广东工业大学学报, 2021, 38(03): 17-21.
[12] 岑仕杰, 何元烈, 陈小聪. 结合注意力与无监督深度学习的单目深度估计[J]. 广东工业大学学报, 2020, 37(04): 35-41.
[13] 曾碧, 任万灵, 陈云华. 基于CycleGAN的非配对人脸图片光照归一化方法[J]. 广东工业大学学报, 2018, 35(05): 11-19.
[14] 陈旭, 张军, 陈文伟, 李硕豪. 卷积网络深度学习算法与实例[J]. 广东工业大学学报, 2017, 34(06): 20-26.
[15] 刘震宇, 李嘉俊, 王昆. 一种基于深度自编码器的指纹匹配定位方法[J]. 广东工业大学学报, 2017, 34(05): 15-21.
Viewed
Full text


Abstract

Cited
  Shared   
  Discussed   
No Suggested Reading articles found!
版权所有 © 2017 《广东工业大学学报》编辑部
地址：广州市东风东路729号广东工业大学 邮编：510090
电话：020-37626653,020-37626139,020-37626396
邮箱：xbzrb@gdut.edu.cn
本系统由北京玛格泰克科技发展有限公司设计开发