Text-to-4D Generation on MAV3D prompts
35.1CLIP Score4D-fy
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| 4D-fyAblation=w/o VL3D (no 3D-aware guidance)2023.11 | 35.1 | 0.88 | 0.89 | 0.95 | 0.92 | 0.91 | |
| 4D-fyAblation=Full pipeline (Ablation Baseline)2023.11 | 35 | — | — | — | — | — | |
| 4D-fyAblation=w/o OLIMG (no VSD guidance)2023.11 | 34.5 | 0.7 | 0.68 | 0.68 | 0.69 | 0.7 | |
| 4D-fyBackbone Architecture=HexPlane2023.11 | 34.5 | 0.95 | 0.92 | 0.9 | 0.92 | 0.95 | |
| 4D-fyEvaluation Context=Quantitative comparison vs MAV3D2023.11 | 34.2 | — | — | — | — | — | |
| MAV3DNote=Metrics for Human Preference represent the win-rate of 4D-fy over this method2023.11 | 33.9 | 0.92 | 0.89 | 0.41 | 0.53 | 0.67 | |
| 4D-fyAblation=w/o hybrid SDS2023.11 | 33.8 | 1 | 1 | 0.78 | 0.88 | 0.95 | |
| 4D-fyAblation=w/o L3D/IMG (no image guidance)2023.11 | 29.3 | 1 | 1 | 0.78 | 0.86 | 0.94 |