Open-vocabulary 3D segmentation on LERF-Mask (test)
92.3Figurines mIoUOP2GS
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| OP2GS2026.05 | 92.3 | 89.1 | 91 | 84.6 | 87 | 85.4 | |
| GagaSegmentation Basis=SAM 2D masks, 2D Segmentation Backbone=Hornet-L2024.04 | 92.3 | 90.8 | 72 | 63.3 | 71.2 | 68.4 | |
| Gaga2026.05 | 90.7 | 89 | 64.1 | 61.6 | 69.3 | 66 | |
| ObjectGS2026.05 | 88.2 | 85.2 | 88 | 79.9 | 88.9 | 88.6 | |
| OmniSeg3DSegmentation Basis=SAM 2D masks2024.04 | 85 | 83.7 | 83.6 | 75.5 | 69.8 | 63.8 | |
| ILGS2026.05 | 75.9 | 73.8 | 84.3 | 75.5 | 81.2 | 78.8 | |
| ILGSSegmentation Basis=SAM 2D masks2024.04 | 75.9 | 73.8 | 81.2 | 78.8 | 84.3 | 75.5 | |
| Gaussian Grouping2026.05 | 69.7 | 67.9 | 77 | 68.7 | 71.7 | 66.1 | |
| Gaussian GroupingSegmentation Basis=SAM 2D masks2024.04 | 69.7 | 67.9 | 77 | 68.7 | 71.7 | 66.1 | |
| GOISegmentation Basis=CLIP feature embedding2024.04 | 63.7 | — | 44.5 | — | 52.6 | — | |
| LangSplatSegmentation Basis=CLIP feature embedding2024.04 | 61.9 | 60.9 | 61.9 | 54.7 | 59.8 | 52.7 | |
| LangSplat2026.05 | 52.8 | 50.5 | 50.4 | 44.7 | 69.5 | 65.6 | |
| DEVA2026.05 | 46.2 | 45.1 | 56.8 | 51.1 | 54.3 | 52.2 | |
| LEGaussiansSegmentation Basis=CLIP feature embedding2024.04 | 34.6 | 32.6 | 31.4 | 18.8 | 42.8 | 35.3 | |
| LERF2026.05 | 33.5 | 30.6 | 28.3 | 14.7 | 49.7 | 42.6 | |
| LERFSegmentation Basis=CLIP feature embedding2024.04 | 33.5 | 30.6 | 28.3 | 14.7 | 49.7 | 42.6 | |
| SA3D2026.05 | 24.9 | 23.8 | 7.4 | 7 | 42.5 | 39.2 | |
| SA3DSegmentation Basis=SAM 2D masks2024.04 | 24.9 | 23.8 | 7.4 | 7 | 42.5 | 39.2 | |
| IGGTSegmentation Basis=SAM 2D masks, Execution=Feed-forward, Max Images per Scene=502024.04 | 17.2 | 12.2 | 5.6 | 0.9 | 21.4 | 18.3 |