Semantic Segmentation on ScanNet (novel views)
64.26mIoUMLP-Splatting
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MLP-Splatting2026.06 | 64.26 | 89.35 | 206 | 12.19 | |
| EPS3DNumber of views=82026.06 | 63.14 | 85.46 | — | — | |
| EPS3DNumber of views=82026.06 | 61.69 | 84.69 | — | — | |
| Feature-3DGS2026.06 | 60.02 | 87 | 1,120 | 0.73 | |
| Uni3RNumber of views=82026.06 | 52.15 | 80.69 | — | — | |
| OursSemantic encoder=CLIP-LSeg, Training views=122025.03 | 50.271 | 83.584 | — | — | |
| Uni3RNumber of views=82026.06 | 48.85 | 80.22 | — | — | |
| LSegNumber of views=82026.06 | 48.43 | 79.77 | — | — | |
| LSegNumber of views=82026.06 | 45.13 | 77.27 | — | — | |
| FSGSSemantic encoder=CLIP-LSeg, Training views=122025.03 | 34.61 | 77.84 | — | — | |
| CoR-GSSemantic encoder=CLIP-LSeg, Training views=122025.03 | 34.17 | 77.944 | — | — | |
| LSMNumber of views=82026.06 | 33.66 | 73.98 | — | — | |
| Gau-GroupingSemantic encoder=CLIP-LSeg, Training views=122025.03 | 33.549 | 73.624 | — | — | |
| LSMNumber of views=82026.06 | 33.51 | 73.5 | — | — | |
| DNGaussianSemantic encoder=CLIP-LSeg, Training views=122025.03 | 32.122 | 74.895 | — | — | |
| SGIOptimization Model=CLIP, Training Views=122025.03 | 28.944 | 68.823 | — | — | |
| 3DOVSSemantic encoder=CLIP-LSeg, Training views=122025.03 | 24.001 | 72.863 | — | — | |
| Feature-3DGSNumber of views=82026.06 | 23.41 | 68.46 | — | — | |
| Feature-3DGSNumber of views=82026.06 | 22.39 | 67.02 | — | — | |
| Feature 3DGSSemantic encoder=CLIP-LSeg, Training views=122025.03 | 21.053 | 72.722 | — | — | |
| Gau-GroupingOptimization Model=CLIP, Training Views=122025.03 | 15.172 | 57.355 | — | — | |
| CoR-GSOptimization Model=CLIP, Training Views=122025.03 | 14.462 | 55.812 | — | — | |
| FSGSOptimization Model=CLIP, Training Views=122025.03 | 14.031 | 56.85 | — | — | |
| DNGaussianOptimization Model=CLIP, Training Views=122025.03 | 13.912 | 54.814 | — | — | |
| 3DOVSOptimization Model=CLIP, Training Views=122025.03 | 7.751 | 36.356 | — | — | |
| Feature 3DGSOptimization Model=CLIP, Training Views=122025.03 | 7.347 | 38.002 | — | — | |
| LangSplatOptimization Model=CLIP, Training Views=122025.03 | 3.339 | 24.491 | — | — |