3D Instance Segmentation on ScanNet200 v1 (val)
30.5mAP@0.50PonderV2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| PonderV2Pre-training=PonderV2, Backbone=PointGroup2023.10 | 30.5 | 37.6 | 20.1 | |
| PointGroup + PPTPre-training=PPT, Backbone=PointGroup2023.10 | 29.4 | 36.8 | 19.4 | |
| PointGroup + MSCPre-training=MSC, Backbone=PointGroup2023.10 | 26.8 | 34.3 | 17.3 | |
| PointGroup + LGroundPre-training=LGround, Backbone=PointGroup2023.10 | 26.1 | — | — | |
| PointGroup + CSCPre-training=CSC, Backbone=PointGroup2023.10 | 25.2 | — | — | |
| PointGroup + PCPre-training=PC, Backbone=PointGroup2023.10 | 24.9 | — | — | |
| PointGroupPre-training=None (Scratch), Backbone=PointGroup2023.10 | 24.5 | 32.2 | 15.8 |