Dataset · codepawl
Cleaned Objaverse 3D models in GLB format
codepawl/objaverse-clean-glb
Curated selection of 1K to 10K engine-ready 3D models under CC0 and CC BY licenses.
What's inside
This dataset contains 1,000 to 10,000 3D objects (along with associated images and text) curated from Objaverse, with broken files and fan art removed. All models are sorted by quality, scaled to a uniform size, and repackaged into web- and game-engine-ready GLB format featuring levels of detail (LOD), a lightweight mobile version, and colliders. Author and license metadata (CC0 1.0 or CC BY 4.0) are preserved in metadata.jsonl and LICENSES.md.
What you can do with it
- Training text-to-3D or image-to-3D generative models on clean geometry
- Populating 3D scenes and simulators with ready-to-use, legally clear assets
- Size
- 1K–10K rows
- Downloads this month
- 7.6K
- Likes
- 23
How to load it
from datasets import load_dataset
ds = load_dataset("codepawl/objaverse-clean-glb")