Novel View Synthesis on DL3DV (in-domain)
25.75PSNRGS-LRM
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GS-LRMResolution=128, Input Views=32, Training Dataset=DL3DV + MegaSynth, Inference Time=0.11 sec2024.12 | 25.75 | 0.859 | 0.13 | |
| Long-LRMResolution=128, Input Views=32, Training Dataset=DL3DV + MegaSynth, Inference Time=0.12 sec2024.12 | 25.44 | 0.853 | 0.136 | |
| Long-LRMResolution=256, Input Views=32, Training Dataset=DL3DV + MegaSynth, Inference Time=0.35 sec2024.12 | 25.14 | 0.828 | 0.186 | |
| GS-LRMResolution=128, Input Views=32, Training Dataset=DL3DV, Inference Time=0.11 sec2024.12 | 24.6 | 0.824 | 0.161 | |
| 3DGSResolution=128, Input Views=32, Training Dataset=N.A. (Per-scene Optimization), Inference Time=5.2 min2024.12 | 24.27 | 0.817 | 0.166 | |
| Long-LRMResolution=128, Input Views=32, Training Dataset=DL3DV, Inference Time=0.12 sec2024.12 | 24.18 | 0.812 | 0.173 | |
| Long-LRMResolution=256, Input Views=32, Training Dataset=DL3DV, Inference Time=0.35 sec2024.12 | 23.71 | 0.779 | 0.236 | |
| 3DGSResolution=256, Input Views=32, Training Dataset=N.A. (Per-scene Optimization), Inference Time=6.4 min2024.12 | 23.26 | 0.778 | 0.206 |