3D Object Detection on Bench2Drive (val)
0.673mAPFlowAD
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| FlowADBackbone=ResNet101, Input Size=512 × 14082026.03 | 0.673 | 0.346 | 0.15 | 0.117 | 0.184 | 0.723 | — | |
| UniDrive-WMparadigm=AR+Diffusion2026.01 | 0.67 | 0.332 | 0.0736 | 0.0258 | 0.247 | 0.752 | — | |
| UniDrive-WMparadigm=Autoregressive (AR)2026.01 | 0.663 | 0.335 | 0.074 | 0.0262 | 0.249 | 0.746 | — | |
| Orion2026.01 | 0.646 | 0.366 | 0.0765 | 0.0271 | 0.258 | 0.723 | — | |
| SparseBEVBackbone=ResNet101, Input Size=512 × 14082026.03 | 0.641 | 0.386 | 0.165 | 0.124 | 0.189 | 0.689 | — | |
| BEVFormerBackbone=ResNet101, Input Size=900 × 16002026.03 | 0.634 | — | — | — | — | 0.671 | — | |
| BEVFormer2026.01 | 0.616 | 0.372 | 0.079 | 0.044 | 0.808 | 0.642 | — | |
| VAD2026.01 | 0.509 | 0.385 | 0.0854 | 0.0308 | 0.594 | 0.604 | — | |
| UniAD2026.01 | 0.121 | 0.518 | 0.17 | 0.096 | 0.977 | 0.316 | — |