3D Visual Grounding on ViGiL3D 2024a (test)
24.1AccuracyGPS
Evaluation results
| Method | Links | ||
|---|---|---|---|
| GPSPretrain=SceneVerse, Finetune=SR, Size=46K2026.06 | 24.1 | 14.7 | |
| V3DMPretrain=ScanScribe, Finetune=ViGiL3D++SR, Size=206K2026.06 | 22.9 | 15.7 | |
| V3DMPretrain=ScanScribe, Finetune=SR, M3D, SR3D, Size=185K2026.06 | 19.2 | 13.6 | |
| ZSVG3D2026.06 | 18.9 | 12.2 | |
| V3DMPretrain=ScanScribe, Finetune=ViGiL3D++, Size=105K2026.06 | 18.6 | 15.5 | |
| V3DMPretrain=ScanScribe, Finetune=SR, Size=46K2026.06 | 17.8 | 11.9 | |
| 3D-GRANDPretrain=3D-GRAND, Finetune=SR, Size=46K2026.06 | 16 | 9.2 | |
| 3D-VisTAPretrain=ScanScribe, Finetune=SR, Size=46K2026.06 | 15.1 | 15.1 | |
| V3DMPretrain=ScanScribe, Finetune=SR, M3D, Size=101K2026.06 | 14.6 | 12.6 |