Open-Vocabulary Segmentation on ADE20K-150 -> ScanNet++ Offices
23W-F1OpenDAS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| OpenDASModality=Both modalities, #Params.=~233K2024.05 | 23 | 30.4 | 21.6 | |
| MaPLeModality=Both modalities, #Params.=~18935K2024.05 | 18.8 | 29.3 | 16.8 | |
| RPOModality=Image-encoder, #Params.=~43K2024.05 | 13.4 | 13.9 | 13.3 | |
| VPTModality=Image-encoder, #Params.=~786K2024.05 | 13 | 19.2 | 8.8 | |
| No adaptationModality=None, #Params.=02024.05 | 11.2 | 11.3 | 11 | |
| CoCoOpModality=Text-encoder, #Params.=~77K2024.05 | 11.2 | 18 | 9.9 |