Semantic View Synthesis on ScanNet V2 (val)
96.8mIoUSemantic-NeRF
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Semantic-NeRFcategorization=Per-scene optimized2023.10 | 96.8 | 98.2 | |
| PNFcategorization=Per-scene optimized2023.10 | 94.1 | 97.2 | |
| DM-NeRFcategorization=Per-scene optimized2023.10 | 93.5 | 96.1 | |
| GNeSFcategorization=Generalizable, training_scale=100 scenes2023.10 | 93.3 | 96.3 | |
| GNeSFcategorization=Generalizable, training_scale=full training set (1000 scenes)2023.10 | 87.8 | 93.3 | |
| Mask2Formercategorization=Generalizable2023.10 | 80.9 | 90.7 | |
| GNeSFcategorization=Generalizable, training_scale=full training set (1000 scenes)2023.10 | 71.6 | 82.3 | |
| Mask2Formercategorization=Generalizable2023.10 | 64.6 | 77.9 | |
| Semantic-Raycategorization=Generalizable, training_scale=full training set2023.10 | 56 | — | |
| Predict Directlycategorization=Generalizable, note=incorporates a semantic head into the generalizable NeRF framework2023.10 | 50.2 | 62.3 | |
| GNeSFcategorization=Generalizable, training_scale=100 scenes2023.10 | 47 | 60.8 | |
| Predict Directlycategorization=Generalizable, note=incorporates a semantic head into the generalizable NeRF framework2023.10 | 39.1 | 54.2 |