Table 2 3D detection on the KITTI dataset
From: Out-of-distribution- and location-aware PointNets for real-time 3D road user detection without a GPU
Method | Cars AP (0.7) | Pedestrians AP (0.5) | Cyclists AP (0.5) | Speed FPS | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
Easy | Moderate | Hard | Easy | Moderate | Hard | Easy | Moderate | Hard | GPU | CPU | |
VoxelNet [3] | 87.99 | 82.47 | 77.34 | 55.35 | 48.79 | 46.22 | 78.44 | 66.40 | 59.20 | 2.3 | 0.1 |
SECOND [4] | 88.21 | 81.32 | 75.23 | 53.35 | 49.66 | 45.79 | 79.53 | 64.69 | 57.32 | 4.7 | 1.2 |
IA-SSD [15] | 89.44 | 81.32 | 76.36 | 45.39 | 41.51 | 37.77 | 79.46 | 62.55 | 57.05 | 19.8 | 4.3 |
PointPillars [9] | 86.64 | 76.74 | 74.16 | 51.46 | 47.94 | 43.80 | 81.86 | 63.66 | 60.91 | 14.5 | 2.1 |
BirdNet [40] | 24.14 | 19.22 | 16.32 | 42.72 | 31.25 | 28.21 | 41.39 | 26.92 | 26.42 | 9.1 | 1.7 |
BirdNet+ [41] | 77.15 | 65.05 | 60.21 | 42.36 | 34.96 | 33.23 | 66.27 | 56.32 | 54.12 | 10.5 | 1.8 |
AVOD [42] | 72.53 | 64.28 | 58.11 | 36.98 | 30.73 | 22.52 | 59.34 | 42.31 | 39.79 | 3.3 | 0.2 |
Ours | 64.34 | 56.25 | 52.01 | 47.41 | 41.12 | 37.61 | 62.85 | 47.12 | 47.26 | 76.1 | 15.2 |