3D Object Detection on V2XSet Perfect setting
88.2AP@0.5V2X-ViT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| V2X-ViTBackbone=PointPillar, Voxel resolution=0.4 m, Compression rate=32, Fusion strategy=Intermediate, Encoder layers=3, MSwin window sizes=4, 8, 162022.03 | 88.2 | 71.2 | |
| V2VNetBackbone=PointPillar, Voxel resolution=0.4 m, Compression rate=32, Fusion strategy=Intermediate2022.03 | 84.5 | 67.7 | |
| DiscoNetBackbone=PointPillar, Voxel resolution=0.4 m, Compression rate=32, Fusion strategy=Intermediate2022.03 | 84.4 | 69.5 | |
| F-CooperBackbone=PointPillar, Voxel resolution=0.4 m, Compression rate=32, Fusion strategy=Intermediate2022.03 | 84 | 68 | |
| Early FusionBackbone=PointPillar, Voxel resolution=0.4 m2022.03 | 81.9 | 71 | |
| OPV2VBackbone=PointPillar, Voxel resolution=0.4 m, Compression rate=32, Fusion strategy=Intermediate2022.03 | 80.7 | 66.4 | |
| Late FusionBackbone=PointPillar, Voxel resolution=0.4 m2022.03 | 72.7 | 62 | |
| No FusionBackbone=PointPillar, Voxel resolution=0.4 m2022.03 | 60.6 | 40.2 |