Feasibility classification on GraspNet-1Billion 1.0 (Novel)
0.996AUROCGRASPFC-PTX
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GRASPFC-PTXEncoder (Architecture)=PTv3, Input=Pose + Cloud, Train. set=MOFEAS-200K2026.06 | 0.996 | 98.5 | 97.1 | |
| GRASPFC-NNETEncoder (Architecture)=PointNeXt-S, Input=Pose + Cloud, Train. set=MOFEAS-200K2026.06 | 0.995 | 97.9 | 96.3 | |
| GRASPFC-PTXEncoder (Architecture)=PTv3, Input=Pose + Cloud, Train. set=MOFEAS-5K-Bal2026.06 | 0.982 | 89.1 | 96.2 | |
| GRASPFC-CONV3DEncoder (Architecture)=VGN + GIGA, Input=Pose + Voxel, Train. set=MOFEAS-200K2026.06 | 0.981 | 97.4 | 91 | |
| GRASPFC-CONV3DEncoder (Architecture)=VGN + GIGA, Input=Pose + Voxel, Train. set=MOFEAS-5K-Bal2026.06 | 0.967 | 91.1 | 92.3 | |
| GRASPFC-CONV3D (no pose)Encoder (Architecture)=VGN, Input=Scene cloud, Train. set=MOFEAS-5K-Bal2026.06 | 0.521 | 42.3 | 61.3 |