3D Object Detection on V2X-Seq
39.9AP (0-150m)Long-SCOPE
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Long-SCOPEBackbone=ResNet50, Number of Queries=900 + 40 (max), Iterative Refinement Layers=42026.04 | 39.9 | 53.2 | 33.9 | 11.3 | |
| Long-SCOPE w/o GQGBackbone=ResNet50, Number of Queries=900, Iterative Refinement Layers=42026.04 | 38.8 | 52.4 | 32.1 | 11.3 | |
| Long-SCOPE w/o CAABackbone=ResNet50, Number of Queries=900 + 40 (max)2026.04 | 35.2 | 53 | 24.3 | 4.1 | |
| SparseCoopBackbone=ResNet50, Number of Queries=9002026.04 | 33.4 | 51.5 | 22.9 | 3.1 | |
| V2X-ViTBackbone=BEVFormer, Grid Resolution=1m x 1m2026.04 | 28.9 | 42.1 | 23.1 | 5.4 | |
| CoopTrackBackbone=ResNet50, Number of Queries=9002026.04 | 23.2 | 32.7 | 19.2 | 5.9 | |
| Late Fusion BaselineBackbone=ResNet502026.04 | 22.4 | 31.8 | 17.8 | 5 | |
| UniV2XBackbone=ResNet50, Number of Queries=9002026.04 | 16.6 | 25.5 | 12.3 | 3.2 | |
| Where2CommBackbone=BEVFormer, Grid Resolution=1m x 1m2026.04 | 5.9 | 14.1 | 2.6 | 0 |