3D Vehicle Detection on Waymo Open Dataset v1.2 (val)
78.12L1 3D mAPOcTr
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| OcTrTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 78.12 | 77.63 | 69.79 | 69.34 | |
| PV-RCNN++Training samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 77.82 | 77.32 | 69.07 | 68.62 | |
| Point2SeqBackbone=Voxel, Head=Sequence2022.03 | 77.52 | 77.03 | 68.8 | 68.36 | |
| PV-RCNNBackbone=Voxel, Head=Anchor2022.03 | 77.51 | 76.89 | 68.98 | 68.41 | |
| PDVTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 76.85 | 76.33 | 69.3 | 68.81 | |
| CenterPointsBackbone=Voxel, Head=Center2022.03 | 76.7 | 76.2 | 68.8 | 68.3 | |
| Pyramid-RCNNBackbone=Voxel, Head=Anchor2022.03 | 76.3 | 75.68 | 67.23 | 66.68 | |
| CT3DBackbone=Voxel, Head=Anchor2022.03 | 76.3 | — | 69.04 | — | |
| SSTTraining samples ratio=20%, IoU threshold=0.7, Training mode=Single-class2023.03 | 76.22 | 75.79 | 68.04 | 67.64 | |
| Voxel R-CNNBackbone=Voxel, Head=Anchor2022.03 | 75.59 | — | 66.59 | — | |
| Voxel-RCNNTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 75.59 | — | 66.59 | — | |
| CenterPointsBackbone=Voxel, Head=Center, re-implemented=true2022.03 | 75.58 | 75.01 | 67 | 66.52 | |
| PV-RCNNTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 75.41 | 74.74 | 67.44 | 66.8 | |
| RSNBackbone=Range, Head=Center2022.03 | 75.1 | 74.6 | 66 | 65.6 | |
| VoTr-TSDBackbone=Voxel, Head=Anchor2022.03 | 74.95 | 74.25 | 65.91 | 65.29 | |
| PartA2NetTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 74.82 | 74.32 | 65.88 | 65.42 | |
| SECONDBackbone=Voxel, Head=Anchor, re-implemented=true2022.03 | 73.62 | 73.11 | 64.86 | 64.4 | |
| LiDAR-RCNNTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 73.5 | 73 | 64.7 | 64.2 | |
| RangeDetBackbone=Range, Head=Center2022.03 | 72.85 | — | — | — | |
| CenterPointTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 71.33 | 70.76 | 63.16 | 62.65 | |
| SECONDTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 70.96 | 70.34 | 62.58 | 62.02 | |
| PointPillarTraining samples ratio=20%, IoU threshold=0.7, Training mode=Multi-class2023.03 | 70.43 | 69.83 | 62.18 | 61.64 | |
| Pillar-ODBackbone=Pillar, Head=Anchor2022.03 | 69.8 | — | — | — | |
| RCDBackbone=Range, Head=Center2022.03 | 69 | 68.5 | — | — | |
| CVCNetBackbone=Voxel, Head=Center2022.03 | 65.2 | — | — | — | |
| AFDetBackbone=Voxel, Head=Center2022.03 | 63.69 | — | — | — | |
| PointPillarsBackbone=Pillar, Head=Anchor2022.03 | 63.3 | 62.7 | 55.2 | 54.7 | |
| MVFBackbone=Voxel, Head=Anchor2022.03 | 62.93 | — | — | — | |
| LaserNetBackbone=Range, Head=Anchor2022.03 | 52.1 | 50.1 | — | — |