• Laser & Optoelectronics Progress
  • Vol. 56, Issue 21, 211004 (2019)
Xujiao Wang, Jie Ma*, Nannan Wang, Pengfei Ma, and Lichaung Yang
Author Affiliations
  • School of Electronics and Information Engineering, Hebei University of Technology, Tianjin 300401, China
  • show less
    DOI: 10.3788/LOP56.211004 Cite this Article Set citation alerts
    Xujiao Wang, Jie Ma, Nannan Wang, Pengfei Ma, Lichaung Yang. Deep Learning Model for Point Cloud Classification Based on Graph Convolutional Network[J]. Laser & Optoelectronics Progress, 2019, 56(21): 211004 Copy Citation Text show less
    Framework of GCN model
    Fig. 1. Framework of GCN model
    Design of point cloud classification network based on GCN
    Fig. 2. Design of point cloud classification network based on GCN
    Schematic of graph convolution in kNN graph layer
    Fig. 3. Schematic of graph convolution in kNN graph layer
    Sampling results of partial shapes on ModelNet40
    Fig. 4. Sampling results of partial shapes on ModelNet40
    OptimizerLearningrateBatchsizeMomentumWeightdecay
    ADAM0.001640.91×10-5
    Table 1. Parameter settings
    ModelMean classaccuracy /%Overallaccuracy /%Forwardtime /ms
    3D ShapeNets[11]77.384.7
    VoxNet[12]8385.9
    PointNet8689.225.3
    ointNet++[13]90.7163.2
    Kd-network[14]86.3
    Ours91.693.229.4
    Table 2. Comparisons of classification accuracy on ModelNet40 and computational complexity of several models
    kMean classaccuracy /%Overallaccuracy /%
    587.689.2
    1088.090.4
    1590.192.6
    2091.693.2
    2591.292.9
    3090.892.5
    3589.991.8
    4089.791.6
    Table 3. Comparison of classification accuracy under different k values
    Xujiao Wang, Jie Ma, Nannan Wang, Pengfei Ma, Lichaung Yang. Deep Learning Model for Point Cloud Classification Based on Graph Convolutional Network[J]. Laser & Optoelectronics Progress, 2019, 56(21): 211004
    Download Citation