Table 2 Results on S3DIS 6-fold dataset for semantic segmentation.
From: EIDU-Net: edge-preserved inception DenseGCN U-Net for LiDAR point cloud segmentation
Method | Input | OA | mAcc | mIoU |
|---|---|---|---|---|
PointNet | Point | 78.5 | 66.2 | 47.6 |
DGCNN | Point | 84.1 | - | 56.1 |
PointCNN | Point | 88.1 | - | 65.4 |
SPGraph | Point | 86.4 | 75.6 | 62.1 |
PAG | Point | 88.1 | - | 65.9 |
PointWeb | Point | 87.3 | 76.2 | 66.7 |
KPConv | Point | - | 79.1 | 70.6 |
RandLA-Net | Point | 88.0 | 82.0 | 70.0 |
DPFA-Net | Point | 89.2 | - | 61.6 |
JSNet++ | Point | 88.7 | 72.8 | 62.4 |
Ours | Point | 88.6 | 75.8 | 63.1 |
