3D Object Detection on Lyft Level 5 mobile objects (test)
82.8AP BEV (0-30m)Sup. (Lyft)
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Sup. (Lyft)Supervision source=Lyft, IoU threshold=0.252022.03 | 82.8 | 82.6 | 70.8 | 70.3 | 50.2 | 49.6 | 69.5 | 69.1 | |
| Sup. (KITTI)Supervision source=KITTI, IoU threshold=0.252022.03 | 79.3 | 78.9 | 57.2 | 56.6 | 30.8 | 29.8 | 58.6 | 57.3 | |
| MODESTSelf-training rounds=R40, IoU threshold=0.252022.03 | 76.4 | 74.1 | 64.2 | 62.9 | 47.1 | 45.5 | 64.4 | 62.7 | |
| MODESTSelf-training rounds=R10, IoU threshold=0.252022.03 | 73.8 | 71.3 | 62.8 | 60.3 | 27 | 24.8 | 57.3 | 55.1 | |
| MODESTSelf-training rounds=R0, IoU threshold=0.252022.03 | 65.7 | 63 | 41.4 | 36 | 8.9 | 5.7 | 42.5 | 37.9 | |
| MODEST-PPSelf-training rounds=R10, IoU threshold=0.25, Condition=Without PP-score-based filtering2022.03 | 49.9 | 49.3 | 32.3 | 27 | 3.5 | 1.4 | 30.9 | 27.3 | |
| MODEST-PPSelf-training rounds=R0, IoU threshold=0.25, Condition=Without multiple traversals2022.03 | 46.4 | 45.4 | 16.5 | 10.8 | 0.9 | 0.4 | 21.8 | 18 |