Fine localization on LangLoc ScanNet 100-scene 1.0 (test)
0.593Positional Error (m) MeanLangLoc w/ dialog
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| LangLoc w/ dialogdialog=true, dialog_source=Qwen2.5-VL-2B2026.07 | 0.593 | 0.532 | 0.21 | 0.069 | 23.77 | 5.12 | 59.3 | |
| LangLocdialog=true2026.07 | 0.593 | 0.532 | 0.21 | 0.069 | 23.77 | 5.12 | 59.3 | |
| Midpointdefinition=floor centroid2026.07 | 1.279 | — | — | 1.098 | — | — | — | |
| Midpointdefinition=floor centroid2026.07 | 1.279 | — | — | 1.098 | — | — | — | |
| VLMbackbone=Qwen2.5-VL-2B, rendering=top-down, zero-shot=true2026.07 | 1.382 | — | — | 1.099 | 93.96 | 91.27 | 3 | |
| VLMmodel=Qwen2.5-VL-2B, rendering=top-down, zero-shot=true2026.07 | 1.382 | — | — | 1.099 | 93.96 | 91.27 | 3 | |
| LangLoc w/o dialogdialog=false2026.07 | 1.676 | 1.254 | 1.065 | 1.314 | 42.67 | 34.66 | 23.6 | |
| LangLocdialog=false2026.07 | 1.676 | 1.254 | 1.065 | 1.314 | 42.67 | 34.66 | 23.6 |