Table 6.
Results of SemanticKITTI validation split. Section 1 shows relevant state-of-the-art methods and their results. Section 2 shows our single projections. The fusion approaches are shown in Section 3 and Section 4. Note that the values differ from the published values. This is because the validation data has been used, and there is no post-processing. For the fusion approach, the NN calculation is included within the inference time, and the architectures are not optimized.
| Method |
Car |
Bicycle |
Motorcycle |
Truck |
Other Vehicle |
Person |
Bicyclist |
Motorcyclist |
Road |
Parking |
Sidewalk |
Other Ground |
Building |
Fence |
Vegetation |
Trunk |
Terrain |
Pole |
Traffic Sign |
mIoU | Mean Time |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RangeNet21 | 0.84 | 0.24 | 0.34 | 0.3 | 0.2 | 0.35 | 0.45 | 0.0 | 0.93 | 0.43 | 0.8 | 0.01 | 0.79 | 0.47 | 0.81 | 0.48 | 0.71 | 0.39 | 0.35 | 0.47 | 76 |
| SalsaNext | 0.86 | 0.39 | 0.42 | 0.78 | 0.42 | 0.62 | 0.68 | 0.0 | 0.94 | 0.42 | 0.8 | 0.04 | 0.80 | 0.48 | 0.80 | 0.58 | 0.64 | 0.47 | 0.44 | 0.55 | 52 |
| PolarNet (cart) | 0.93 | 0.27 | 0.50 | 0.47 | 0.27 | 0.53 | 0.72 | 0.0 | 0.89 | 0.36 | 0.72 | 0.95 | 0.89 | 0.46 | 0.86 | 0.6 | 0.74 | 0.53 | 0.43 | 0.53 | 39 |
| Bird’s-eye | 0.8 | 0.04 | 0.11 | 0.06 | 0.12 | 0.17 | 0.36 | 0.0 | 0.83 | 0.22 | 0.62 | 0.92 | 0.85 | 0.17 | 0.62 | 0.37 | 0.65 | 0.22 | 0.19 | 0.34 | 33 |
| 3D bird’s-eye | 0.92 | 0.11 | 0.25 | 0.64 | 0.3 | 0.32 | 0.69 | 0.0 | 0.88 | 0.34 | 0.69 | 0.0 | 0.86 | 0.35 | 0.82 | 0.52 | 0.7 | 0.47 | 0.26 | 0.48 | 49 |
| 3D bird’s-eye | 0.9 | 0.2 | 0.68 | 0.53 | 0.18 | 0.45 | 0.72 | 0.0 | 0.89 | 0.41 | 0.71 | 0.0 | 0.89 | 0.47 | 0.86 | 0.58 | 0.74 | 0.51 | 0.4 | 0.5 | 114 |
| 3D cylindrical | 0.84 | 0.1 | 0.1 | 0.1 | 0.26 | 0.25 | 0.25 | 0.0 | 0.84 | 0.17 | 0.63 | 0.0 | 0.77 | 0.32 | 0.79 | 0.45 | 0.66 | 0.39 | 0.23 | 0.37 | 47 |
| SalsaNext [3 × 64 × 512] | 0.89 | 0.34 | 0.52 | 0.76 | 0.46 | 0.47 | 0.5 | 0.0 | 0.93 | 0.45 | 0.78 | 0.0 | 0.77 | 0.5 | 0.8 | 0.5 | 0.67 | 0.34 | 0.4 | 0.53 | 11 |
| Baseline | 0.91 | 0.32 | 0.45 | 0.81 | 0.38 | 0.42 | 0.64 | 0.0 | 0.91 | 0.41 | 0.75 | 0.0 | 0.88 | 0.51 | 0.85 | 0.55 | 0.71 | 0.46 | 0.34 | 0.54 | 240 |
| KPFusion | 0.93 | 0.35 | 0.26 | 0.62 | 0.4 | 0.38 | 0.65 | 0.0 | 0.89 | 0.37 | 0.72 | 0.0 | 0.88 | 0.51 | 0.86 | 0.59 | 0.72 | 0.55 | 0.46 | 0.53 | 386 |
| PointNetFusion | 0.91 | 0.06 | 0.17 | 0.66 | 0.36 | 0.2 | 0.09 | 0.0 | 0.89 | 0.34 | 0.71 | 0.0 | 0.86 | 0.41 | 0.83 | 0.36 | 0.7 | 0.32 | 0.35 | 0.43 | 291 |
| NN | 0.92 | 0.32 | 0.47 | 0.8 | 0.4 | 0.44 | 0.66 | 0.0 | 0.91 | 0.42 | 0.75 | 0.0 | 0.88 | 0.5 | 0.85 | 0.56 | 0.71 | 0.49 | 0.35 | 0.55 | 327 |
| KPFusion | 0.94 | 0.4 | 0.42 | 0.73 | 0.42 | 0.67 | 0.8 | 0.0 | 0.94 | 0.43 | 0.8 | 0.0 | 0.9 | 0.59 | 0.89 | 0.7 | 0.78 | 0.59 | 0.49 | 0.61 | 445 |
| PointNetFusion | 0.9 | 0.19 | 0.08 | 0.3 | 0.22 | 0.51 | 0.7 | 0.0 | 0.91 | 0.36 | 0.76 | 0.2 | 0.88 | 0.5 | 0.87 | 0.61 | 0.76 | 0.57 | 0.45 | 0.5 | 342 |
Car
Bicycle
Motorcycle
Truck
Other Vehicle
Person
Bicyclist
Motorcyclist
Road
Parking
Sidewalk
Other Ground
Building
Fence
Vegetation
Trunk
Terrain
Pole
Traffic Sign