3D Object Classification on Roadside LiDAR 30-shot 1.0 (test)
0.705F1 Score @ 30PointNet
Evaluation Results
| Method | Links | |
|---|---|---|
| PointNetInput Representation=Fused 3D Point Cloud, Training status=Yes, Adapt. Time (s)=198.3, Infer. Time (ms)=4.18 ± 0.212026.02 | 0.705 | |
| Ours (Linear Probe)Adaptation Strategy=Linear Probe, Input Representation=Proposed 2D Proxy, Training status=Yes, Adapt. Time (s)=12.4, Infer. Time (ms)=19.54 ± 0.302026.02 | 0.649 | |
| Ours (Few-Shot Learning)Adaptation Strategy=Few-Shot Learning, Input Representation=Proposed 2D Proxy, Training status=No, Adapt. Time (s)=11.4, Infer. Time (ms)=19.76 ± 0.252026.02 | 0.621 | |
| ViT-B/16Input Representation=Proposed 2D Proxy, Training status=Yes, Adapt. Time (s)=300.4, Infer. Time (ms)=8.97 ± 0.322026.02 | 0.087 |