Novel View Synthesis on Mip-NeRF 360 (Indoor)
31.16PSNRGS^2
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GS^22026.04 | 31.16 | 93 | 0.18 | 0.27 | — | |
| 3DGS2026.04 | 31.09 | 92 | 0.19 | 1.41 | — | |
| MaskGaussian2026.04 | 30.91 | 93 | 0.19 | 0.51 | — | |
| Mini-Splatting2026.04 | 30.61 | 92 | 0.19 | 0.4 | — | |
| Compact 3DGS2026.04 | 30.54 | 92 | 0.18 | 1.21 | — | |
| EAGLES2026.04 | 30.46 | 93 | 0.18 | 0.71 | — | |
| LightGaussian2026.04 | 30.41 | 92 | 0.19 | 0.94 | — | |
| 3DGS#G ×10^6=1.45, Storage (MB)=478, Train (min)=272026.06 | 30.41 | — | — | — | — | |
| Mobile-GS#G ×10^6=0.38, Storage (MB)=3.5, Train (min)=862026.06 | 30.37 | — | — | — | 131 | |
| CompGS2026.04 | 30.33 | 92 | 0.19 | 0.46 | — | |
| Flux-GS#G ×10^6=0.22, Storage (MB)=2.1, Train (min)=112026.06 | 30.22 | — | — | — | 147 | |
| Speedy-Splat#G ×10^6=0.37, Storage (MB)=64, Train (min)=142026.06 | 30.11 | — | — | — | 21 | |
| LocoGS-S#G ×10^6=0.82, Storage (MB)=6.1, Train (min)=462026.06 | 30.08 | — | — | — | 13 | |
| 3DGS*Quantization=Huffman encoding, #G ×10^6=1.45, Storage (MB)=46, Train (min)=272026.06 | 30.04 | — | — | — | 11 | |
| C3DGS#G ×10^6=0.75, Storage (MB)=21, Train (min)=312026.06 | 30.01 | — | — | — | 18 | |
| Mobile-GS*Inference Stage=Without MLP, #G ×10^6=0.38, Storage (MB)=3.4, Train (min)=642026.06 | 29.58 | — | — | — | 142 |