USDCraft: Geometrically Grounded Programmatic Modeling of Articulated 3D Assets for Simulation
Paper • 2610.11322 • Published • 18
Dataset Viewer
The dataset could not be loaded because the splits use different data file formats, which is not supported. Read more about the splits configuration. Click for more details.
Couldn't infer the same data file format for all splits. Got {NamedSplit('train'): ('json', {}), NamedSplit('test'): ('meshfolder', {})}
Error code: FileFormatMismatchBetweenSplitsError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Object Library web models
Compact GLB versions of free 3D asset collections, for viewing in the browser.
Browse them all in 3D at objects.geronimo-labs.com — search, rotate any asset, save a view as a PNG or download its GLB.
| Collection | Assets | Folder | Source | License |
|---|---|---|---|---|
| NVIDIA SimReady | 2,677 | <asset-id>/ |
nvidia/simready-assets | CC BY 4.0, © NVIDIA |
| Google Scanned Objects | 1,033 | google-scanned-objects/<name>/ |
GoogleResearch on Gazebo Fuel | CC BY 4.0, © 2020 Google LLC |
| NVIDIA USDCraft-10K | 10,000 | usdcraft-10k/<name>/ |
nvidia/USDCraft-10K | Apache 2.0, © NVIDIA |
NVIDIA SimReady
Each <asset-id>/model.glb holds the visible geometry with WebP textures (at most 1024 px) and Draco-compressed meshes, converted from the original USD files. Physics data is not included; use the original dataset for simulation.
Google Scanned Objects
Photo-scanned household items (shoes, toys, packaged goods and more). Each google-scanned-objects/<name>/ holds:
model.glb— the scanned mesh with its texture as WebP (at most 2048 px), Draco-compressed, converted from the original OBJthumb.png— a transparent 768 px rendermeta.json— name, brand, categories, copyright and license from the original model
Many items are branded retail products; the scans are CC BY 4.0, but the brands and logos shown on them belong to their owners.
Reference: Downs et al., Google Scanned Objects: A High-Quality Dataset of 3D Scanned Household Items, ICRA 2022.
NVIDIA USDCraft-10K
AI-generated, simulation-ready objects: appliances, furniture, tools, lab and sports gear. For each object, a language model (DeepSeek-V4.1-Flash) wrote a program from a text prompt, which NVIDIA's USDCraft compiled into an articulated USD asset with parts, joints and physics. Each usdcraft-10k/<name>/ holds:
model.glb— the visible geometry, Draco-compressed, converted from the original USD. Joints and physics are not included.thumb.png— a transparent 768 px rendermeta.json— category, prompt, parts and joints from the original dataset
Reference: USDCraft: Geometrically Grounded Programmatic Modeling of Articulated 3D Assets for Simulation, arXiv:2610.11322.
License
SimReady and Google files are CC BY 4.0, USDCraft-10K files Apache 2.0. When you use them, credit the original authors: NVIDIA for SimReady and USDCraft-10K, Google for Google Scanned Objects.
- Downloads last month
- 4,552