Text-guided 3D synthesis on Manually created dataset of diverse text prompts and objects
85.29CLIP R-Precision (ViT-B/32)Dream3D
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Dream3DGuiding model=CLIP ViT-B/16, Optimization steps=5000, 3D prior=true2022.12 | 85.29 | 98.53 | |
| Dream3D w/o 3D priorGuiding model=CLIP ViT-B/16, Optimization steps=5000, 3D prior=false2022.12 | 75.47 | 94.34 | |
| CLIP-MeshGuiding model=CLIP ViT-B/162022.12 | 75 | 91.18 | |
| PureCLIPNeRFGuiding model=CLIP ViT-B/16, Optimization steps=>=10000, Variant=implicit architecture2022.12 | 73.53 | 88.24 | |
| DreamFieldsGuiding model=CLIP ViT-B/16, Optimization steps=>=10000, Configuration=medium-quality2022.12 | 63.24 | 92.65 |