Table 2 Detailed semantic segmentation results for projected image datasets
From: Cross modal networks for point cloud semantic segmentation of Chinese ancient buildings
Methods | mIou (%) | aAcc (%) | Per Class IoU(%) | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
bm | cn | dr | df | fl | lt | or | pl | pq | rf | tb | tr | wl | wd | |||
DeepLabV356 | 22.18 | 69.63 | 24.0 | 0 | 19.34 | 12.83 | 63.24 | 16.65 | 2.91 | 37.35 | 0 | 62.6 | 9.21 | 3.45 | 57.43 | 1.45 |
CCNet57 | 22.94 | 70.02 | 25.91 | 0 | 25.55 | 9.02 | 62.98 | 19.45 | 1.71 | 36.47 | 0 | 63.77 | 9.46 | 6.51 | 59.48 | 0.86 |
FCN58 | 23.63 | 71.6 | 26.01 | 1.23 | 24.34 | 13.35 | 65.86 | 18.69 | 2.61 | 38.76 | 0 | 64.64 | 9.6 | 4.23 | 60.67 | 0.78 |
DeepLabV3 + 59 | 24.91 | 72.17 | 28.51 | 0 | 24.97 | 16.38 | 69.63 | 17.72 | 3.56 | 39.1 | 0 | 65.73 | 11.74 | 0.71 | 61.43 | 9.32 |
GCNet60 | 69.89 | 90.55 | 65.08 | 72.44 | 75.74 | 77.94 | 89.7 | 62.57 | 41.07 | 72.8 | 55.18 | 84.19 | 55.83 | 70.4 | 87.81 | 67.67 |
SegFormer61 | 70.88 | 90.71 | 65.51 | 71.71 | 75.30 | 77.44 | 89.67 | 64.45 | 40.46 | 73.46 | 63.83 | 84.60 | 57.48 | 71.40 | 88.1 | 68.9 |
Mask2Former55 | 74.31 | 91.30 | 68.59 | 79.80 | 79.44 | 81.59 | 90.10 | 68.86 | 45.84 | 76.63 | 69.54 | 85.11 | 61.91 | 73.63 | 88.73 | 70.60 |
