Use case | Custom datasets
Tell us about your use case. We'll build the dataset — and the endpoint to grow it.
Send a brief. Get a custom dataset of sim-ready 3D assets, plus a private API
endpoint to generate more on demand.
How it works
How it works
Step 01
luggage-01.jpg2.1MB
DimensionsXX
SpecsXY
DetailYZ
Step 02
luggage-01.jpg2.1MB
DimensionsXX
SpecsXY
DetailYZ
Step 03
luggage-01.jpg2.1MB
DimensionsXX
SpecsXY
DetailYZ
custom_run
"density_kg_m3": 251.6274268722595,"static_friction": 0.65,"dynamic_friction": 0.62,"restitution": 0.21,"units": "SI","body_type": "rigid","surface_material": "nylon_woven","mass_kg": 2.4,"collider": "sdf","validation": "drop_test_pass","asset_id": "luggage-01-a3f","usd": "s3://datasets/luggage/01.usd"
Why custom
Why it’s built this way
Scoped to your objects
The exact classes, body types (rigid, deformable, articulated), and conditions your robot meets, not a generic library.
Domain randomization tuned to your dataset
The variation axes are set per object class in your brief: geometry, texture, and the physics properties that actually vary for your items (mass for variable-content containers, deformable properties for soft goods). What doesn't vary in reality is calculated, not randomized.
A private endpoint, not a hand-off
Regenerate and extend the dataset on demand, without re-briefing.
Sim-ready throughout
Every asset carries full physics tags, a dedicated collision mesh, and a drop-test validation report, delivered in USD for Isaac Sim, Isaac Lab, and MuJoCo.
Built for
Teams with specific coverage requirements — a defined set of objects and conditions to validate against — who need that set to grow with their program instead of stalling on manual sourcing.
