3D Object Detection on Waymo Open Dataset (WOD) (val) (Granular 3D AP)
57.79Vehicle 3D AP L1 (IoU=0.5)CPD
Evaluation results
| Method | Links | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| CPDsupervision=unsupervised, backbone=CenterPoint-like2024.04 | 57.79 | 37.4 | 50.18 | 32.13 | 21.91 | 16.31 | 18.01 | 13.22 | 5.83 | 5.06 | 5.61 | 4.87 | |
| Proto-vanillasupervision=unsupervised2024.04 | 35.22 | 20.19 | 31.58 | 18.36 | 17.6 | 10.34 | 14.62 | 8.59 | 4.21 | 3.45 | 3.8 | 3.31 | |
| OYSTERsupervision=unsupervised2024.04 | 30.48 | 14.66 | 26.21 | 14.1 | 4.33 | 0.18 | 3.52 | 0.14 | 1.27 | 0.33 | 1.24 | 0.32 | |
| MODESTsupervision=unsupervised2024.04 | 18.51 | 6.46 | 15.83 | 5.48 | 11.83 | 0.17 | 8.96 | 0.1 | 1.47 | 1.14 | 1.17 | 1.01 | |
| DBSCAN init-trainsupervision=unsupervised2024.04 | 17.36 | 2.65 | 14.87 | 2.29 | 1.65 | 0 | 1.35 | 0 | 0.48 | 0.25 | 0.43 | 0.2 | |
| DBSCANsupervision=unsupervised2024.04 | 2.32 | 0.29 | 1.94 | 0.25 | 0.51 | 0 | 0.19 | 0 | 0.28 | 0.03 | 0.2 | 0 |