Table 2 Performance comparison with LIDAR-based and sensor fusion-based methods for vehicle detection on the KITTI benchmark (AP).
From: Fast vehicle detection based on colored point cloud with bird’s eye view representation
Methods | Sensor modality | Time (s) | 3D detection (%) | 2D BEV detection (%) | ||||
|---|---|---|---|---|---|---|---|---|
Easy | Moderate | Hard | Easy | Moderate | Hard | |||
BirdNet5 | L (BEV) | 0.11 | 40.99 | 27.26 | 25.32 | – | – | – |
BirdNet+6 | L (BEV) | 0.1 | 70.14 | 51.85 | 50.03 | 81.85 | 86.43 | 75.36 |
RT3D8 | L (BEV) | 0.09 | 23.74 | 19.14 | 18.86 | 56.44 | 44.00 | 42.34 |
Pointpillars10 | L (BEV) | 0.016 | 82.58 | 74.31 | 68.99 | 88.35 | 86.10 | 79.83 |
VoxelNet16 | L (voxel) | 0.23 | 77.47 | 65.11 | 57.73 | 89.35 | 79.26 | 77.39 |
SECOND17 | L (voxel) | 0.04 | 84.65 | 75.96 | 68.71 | 88.07 | 79.37 | 77.95 |
SegVoxelNet18 | L (voxel) | 0.04 | 86.04 | 76.13 | 70.76 | 86.62 | 86.16 | 78.68 |
MV3D25 | L + C | 0.36 | 74.97 | 63.63 | 54.00 | 85.82 | 76.90 | 68.94 |
AVOD26 | L + C | 0.08 | 76.39 | 66.47 | 60.23 | 86.80 | 83.79 | 77.90 |
MVAF-Net29 | L + C | 0.06 | 87.87 | 78.71 | 75.48 | – | – | – |
MMF30 | L + C | 0.08 | 88.40 | 77.43 | 70.22 | 89.49 | 87.47 | 79.10 |
Contfuse38 | L + C | 0.06 | 82.54 | 66.22 | 64.04 | 88.81 | 85.83 | 77.33 |
F-PointNet40 | L + C | 0.17 | 82.19 | 69.79 | 60.59 | 88.70 | 84.00 | 75.33 |
Ours | L + C | 0.05 | 89.14 | 77.85 | 73.03 | 91.74 | 85.26 | 79.73 |
