3D Scenes
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| 3D Editing Defense | 3D Scenes evaluated via GaussianEditor (test) | CLIP Original Score1 | 5 | |
| Text-guided 3D scene generation | 3D Scenes with Qwen1.5 captions (Scene chunks) | CLIP-Score23.79 | 4 | |
| Text-guided 3D scene generation | 3D Scenes with Qwen1.5 captions (Independent chunks) | CLIP-Score23.96 | 4 | |
| 3D Scene Generation | 3D Scenes average across all scene types | Avg FID24.34 | 3 | |
| Active View Planning | Large 3D Scenes (test) | Dunnottar Castle73.9 | 3 | |
| Image Spatial Reasoning | 3D Scenes (3 outdoor and 1 indoor scenes) | SR67 | 2 |
Showing 6 of 6 rows