3D Object Detection on KITTI (val) (Distance Bins)
88AP BEV (0-30m)Supervised (KITTI)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Supervised (KITTI)Training rounds=last, Training dataset=KITTI, Supervision source=Ground-truth, IoU threshold=0.252022.03 | 88 | 87.9 | 73.6 | 72 | 46.7 | 45.4 | 81.1 | 81 | |
| Supervised (Lyft)Training rounds=last, Training dataset=Lyft, Supervision source=Ground-truth, IoU threshold=0.252022.03 | 82 | 81.9 | 53.4 | 51.8 | 24.9 | 22.2 | 71.3 | 69.6 | |
| MODESTTraining rounds=40, Training dataset=Lyft, Supervision source=Self-labeled, IoU threshold=0.252022.03 | 73.6 | 73.2 | 49.9 | 48.2 | 15.1 | 13.9 | 63 | 61.1 | |
| MODESTTraining rounds=10, Training dataset=Lyft, Supervision source=Self-labeled, IoU threshold=0.252022.03 | 73.5 | 71.6 | 50.3 | 48.3 | 9.6 | 8.1 | 61.7 | 59.7 | |
| MODEST-PPTraining rounds=10, Training dataset=Lyft, Supervision source=Self-labeled (without PP-score filtering), IoU threshold=0.252022.03 | 56.6 | 55 | 22.7 | 18.1 | 0.9 | 0.7 | 42.7 | 40.7 |