Apppublic

rigyd/ycb-simready

sourceHugging Facecc-by-4.0updated 8d agoView on Hugging Face

2likes

App README

Rigyd × YCB: SimReady sample

Five objects from the YCB Object and Model Set as simulation-ready assets for robot learning. Rigyd converted each one, and each ships with:

—OpenUSD. A composed SimReady USD tree (usd/<name>.usd + payloads/). It passes the SimReady Prop-Robotics Neutral, PhysX and Isaac profiles (spec 2026.06.0) with 0 errors.
—MuJoCo. An MJCF (mjcf/<name>.xml) with a textured visual mesh and convex collision hulls.
—Physics. Mass, full inertia tensor, centre of mass, friction and restitution, identical in both files. Mass is compared against the published YCB mass.
—Report. report.json records where every physical value came from.

Object	YCB mass	Rigyd mass
005 Tomato Soup Can	349 g	340 g
006 Mustard Bottle	603 g	600 g
025 Mug	118 g	113 g
048 Hammer	665 g	680 g
056 Tennis Ball	58 g	57.7 g

Every object is upright (Z-up, metres), keeps the exact size of the scan, and comes to rest when dropped onto a plane in MuJoCo.

Download

Use Download USD + MJCF on any asset, or fetch everything:

python

from huggingface_hub import snapshot_download

folder = snapshot_download(repo_id="rigyd/ycb-simready", repo_type="space", allow_patterns=["assets/**"])

Open assets/YCB/<name>/usd/<name>.usd in Isaac Sim or usdview. Keep the whole usd/ folder: the root layer composes payloads/. Load assets/YCB/<name>/mjcf/<name>.xml in MuJoCo.

License

The YCB scans and these derived assets are licensed CC BY 4.0 (Calli et al., The YCB Object and Model Set, ICAR 2015); see assets/LICENSE.txt. Product labels visible in the textures belong to their owners.

The viewer is built on the MIT-licensed USD Asset Library template (see LICENSE). Rigyd's changes:

—textured previews;
—one entry per composed asset;
—per-asset physics;
—USD + MJCF package downloads.

More at rigyd.com.