hi3d-official/Hi3Dv1.0
AI model · https://huggingface.co/hi3d-official/Hi3Dv1.0
First seen by Oriel on .
AI summary
Hi3D v1.0 refines a coarse mesh from a reference image into high-resolution geometry, using Sparc3D-based progressive sparse sampling and extraction at effective resolutions up to 1536³.
- The model was trained on approximately two million curated 3D assets and uses three passes with 512³, 1024³, and 1536³ sparse extraction grids.
- Inference requires Linux, Python 3.10, CUDA 12.0–12.5, and an NVIDIA GPU with compute capability 8.0 or newer; no CPU path exists.
- The verified 1536³ setup used a 96 GB NVIDIA H20, while feasibility and peak memory on smaller GPUs remain unestablished.
- Outputs contain geometry only, use batch size one, discard world coordinates and absolute scale, and may preserve severe structural errors from the coarse mesh.
- Training code, evaluation tooling, benchmark splits, and comparison meshes are excluded, so the repository cannot reproduce the report’s comparative tables.
#3d geometry #sparc3d #sparse vae #computer vision
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Models: 146, latest observation ; +32 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Models: 0, latest observation ; 0 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Models | hi3d-official/Hi3Dv1.0 | https://huggingface.co/hi3d-official/Hi3Dv1.0 Read discussion |