Novel View Synthesis on Google Scanned Objects (GSO) (test)
27.522PSNRLVSM
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| LVSMPose Input=posed2026.03 | 27.522 | 0.902 | — | 0.0895 | |
| RnGPose Input=unposed2026.03 | 26.276 | 0.891 | — | 0.0975 | |
| SODAApproach=Diffusion-based (Geometry-Free)2023.11 | 24.97 | 0.945 | 1.51 | 0.054 | |
| PixelNeRFApproach=Geometry-Aware2023.11 | 24.93 | 0.919 | 48.72 | 0.086 | |
| Vanilla Auto-EncoderApproach=Geometry-Free2023.11 | 24.46 | 0.941 | 75.96 | 0.129 | |
| SODA (shorter sampling)Sampling steps=20, Approach=Diffusion-based (Geometry-Free)2023.11 | 24.38 | 0.93 | 2.35 | 0.065 | |
| NeRF-VAEApproach=Geometry-Aware2023.11 | 22.2 | 0.882 | 74.83 | 0.113 | |
| SRTApproach=Geometry-Free2023.11 | 21.97 | 0.877 | 38.64 | 0.11 | |
| SPARSEGENResolution=128 x 128, Training Dataset=Objaverse, Inference speed=0.033s, Representation size=560KB2026.04 | 21.427 | 0.85 | — | 0.16 | |
| Splatter ImageResolution=128 x 128, Training Dataset=Objaverse2026.04 | 21.065 | 0.878 | — | 0.111 | |
| SODA w/o modulationFeature modulation=false, Approach=Diffusion-based (Geometry-Free)2023.11 | 21.02 | 0.929 | 4.1 | 0.069 | |
| SODA w/o bottleneckBottleneck=false, Approach=Diffusion-based (Geometry-Free)2023.11 | 20.97 | 0.926 | 4.39 | 0.071 | |
| SyncDreamer + DreamComposercondition_view=generated from diffusion models2023.12 | 20.52 | 0.828 | — | 0.141 | |
| SyncDreamercondition_view=generated from diffusion models2023.12 | 20.05 | 0.798 | — | 0.146 | |
| Matrix3DPose Input=posed2026.03 | 19.941 | 0.803 | — | 0.1684 | |
| VGGT + LVSMPose Input=unposed2026.03 | 19.394 | 0.796 | — | 0.198 | |
| Zero-1-to-3condition_view=generated from diffusion models2023.12 | 18.93 | 0.779 | — | 0.166 | |
| Matrix3DPose Input=unposed2026.03 | 18.736 | 0.786 | — | 0.193 | |
| LGMPose Input=posed2026.03 | 16.677 | 0.751 | — | 0.2248 | |
| DALL-E 2Approach=Diffusion-based (Geometry-Free)2023.11 | 15.68 | 0.793 | 8.32 | 0.147 | |
| Realfusioncondition_view=generated from diffusion models2023.12 | 15.26 | 0.722 | — | 0.283 | |
| OpenLRMResolution=128 x 128, Training Dataset=Objaverse2026.04 | 14.526 | 0.741 | — | 0.199 | |
| PaletteApproach=Diffusion-based (Geometry-Free)2023.11 | 13.42 | 0.672 | 8.49 | 0.199 | |
| pixelSplatPose Input=posed2026.03 | 12.904 | 0.718 | — | 0.4514 |