Novel View Generation on VIGOR OOD (test)
19.2FIDSat3DGen
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Sat3DGenglobal illumination feature=randomly selected from training set2026.05 | 19.2 | 0.014 | 0.525 | 0.37 | 12.83 | 0.38 | 0.3 | |
| ControlNettype=diffusion-based image generation model2026.05 | 23.6 | — | — | 0.34 | 12.02 | 0.46 | 0.34 | |
| ControlS2Stype=diffusion-based image generation model2026.05 | 28 | — | — | 0.42 | 13.8 | 0.38 | 0.27 | |
| Canonical Image-to-3Darchitecture=feed-forward image-to-3D model, configuration=removes proposed spatial token module, gravity-based density variation loss, perspective training strategy2026.05 | 35.6 | 0.03 | 0.479 | 0.35 | 12.63 | 0.42 | 0.32 | |
| Sat2Density++global illumination feature=randomly selected from training set2026.05 | 40.8 | 0.035 | 0.465 | 0.34 | 12.51 | 0.44 | 0.34 | |
| Sat2Densityglobal illumination feature=randomly selected from training set2026.05 | 85.6 | 0.079 | 0.451 | 0.32 | 12.48 | 0.45 | 0.37 |