Table 4 Comparative experiments on reconstruction quality of different models.
From: Generation driven understanding of localized 3D scenes with 3D diffusion model
Method | PSNR(dB)\(\uparrow\) | SSIM\(\uparrow\) | LPIPS\(\downarrow\) | IoU\(\uparrow\) | Chamfer distance (CD)\(\downarrow\) | F-score\(\uparrow\) |
|---|---|---|---|---|---|---|
\(\pi\)-GAN | 28.65 | 0.924 | 0.121 | 0.807 | 0.030 | 0.88 |
EG3D | 27.01 | 0.919 | 0.180 | 0.861 | 0.035 | 0.85 |
DiffRF | 25.83 | 0.887 | 0.210 | 0.738 | 0.045 | 0.82 |
AutoSDF | 28.78 | 0.879 | 0.113 | 0.875 | 0.032 | 0.86 |
3D-UDDPM (ours) | 29.83 | 0.917 | 0.090 | 0.912 | 0.022 | 0.91 |
