3D Referring Expression Segmentation on ScanRefer Multiple
0.55Acc@25RG-SAN
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RG-SANInput Modality=ground-truth point clouds2026.01 | 0.55 | 0.354 | 0.374 | |
| Reason3DInput Modality=ground-truth point clouds2026.01 | 0.505 | 0.317 | 0.341 | |
| MVGGTInput Modality=sparse RGB inputs2026.01 | 0.492 | 0.335 | 0.338 | |
| 3D-STMNInput Modality=ground-truth point clouds2026.01 | 0.462 | 0.292 | 0.311 | |
| TGNNInput Modality=ground-truth point clouds2026.01 | 0.312 | 0.266 | 0.236 | |
| two-stageInput Modality=sparse RGB inputs2026.01 | 0.251 | 0.122 | 0.164 | |
| 2D-LiftInput Modality=sparse RGB inputs2026.01 | 0.215 | 0.056 | 0.145 |