基于边界点估计与稀疏卷积神经网络的三维点云语义分割
杨军,张琛

Semantic segmentation of 3D point cloud based on boundary point estimation and sparse convolution neural network
Jun YANG,Chen ZHANG
表 1 不同方法在S3DIS数据集上的分割精度对比(以Area 5作为测试)
Tab.1 Comparison of segmentation accuracy of different methods on S3DIS dataset (Area 5 as a test)
方法OA/%mIoU/%IoU/%
ceilingfloorwallbeamcolumnwindowdoortablechairsofabookcaseboardclutter
PointNet [5]79.341.188.897.369.80.13.946.310.859.052.65.940.326.433.2
TangentConv [28]82.552.690.597.774.00.020.739.031.377.569.457.338.548.839.8
PointCNN [29]85.957.392.398.279.40.017.622.862.174.480.631.766.762.156.7
SPG [30]86.458.089.496.978.10.042.848.961.684.775.469.852.62.152.2
PointWeb [31]87.060.392.098.579.40.021.159.734.876.388.346.969.364.952.5
HPEIN [32]87.261.991.598.281.40.023.365.340.075.587.758.567.865.649.4
RandLA-Net [18]87.262.491.195.680.20.024.762.347.776.283.760.271.165.753.8
GACNet [33]87.862.892.398.381.90.020.359.140.878.585.861.770.774.752.8
PPCNN++ [34]64.094.098.583.70.018.666.161.779.488.049.570.166.456.1
BAAF-Net [35]88.965.492.997.982.30.023.165.564.978.587.561.470.768.757.2
KPConv [36]67.192.897.382.40.023.958.069.081.591.075.475.366.758.9
AGConv [37]90.067.993.998.482.20.023.959.171.391.581.275.574.972.158.6
本研究方法90.869.594.499.287.20.027.262.272.891.885.879.066.774.462.9