Datasets:

thisisrick25
/

nvidia-simready-catalog

license: cc-by-4.0
task_categories:
  - robotics
tags:
  - robotics
  - simulation
  - OpenUSD
  - Isaac-Sim
  - synthetic-data
  - physics
  - mujoco
size_categories:
  - 1K<n<10K
pretty_name: SimReady Assets (NVIDIA)

NVIDIA SimReady Assets — Community Dataset Card

This card was produced by the community catalog project and offered to NVIDIA; numbers are computed programmatically from the repo tree and payload files on 2026-11-04. Verify against https://huggingface.co/datasets/nvidia/simready-assets for the latest state.

NVIDIA's SimReady Assets are simulation-ready 3D models for robotics and Physical AI: every asset ships with tuned physics (mass, inertia, collision approximations), materials, and variants for PhysX, MuJoCo, and Newton physics engines. The dataset contains 2,677 assets totaling 489.68 GB under CC-BY-4.0.

Dataset at a glance


Assets	2,677
Total size	489.68 GB
Files	183,398
Categories	14
Articulated assets	5 (0.2%)
PhysX variant	2,669
MuJoCo variant	2,669
Newton variant	2,669
License	CC-BY-4.0
Data as of	2026-11-04

Directory structure

Each asset lives in its own top-level directory:

<asset_name>/
├── com.nvidia.simready.packaging.json   # package id, license, sha256 hashes
├── .metadata/                           # bill of materials
├── simready_usd/
│   ├── sm_<asset_name>_01.usd           # main SimReady USD (binary USDC)
│   ├── payloads/
│   │   ├── base.usda                    # structure + variant sets
│   │   ├── instances.usda               # per-mesh physics (neutral)
│   │   ├── instances_physx.usda          # per-mesh physics (PhysX)
│   │   ├── instances_mujoco.usda         # per-mesh physics (MuJoCo)
│   │   ├── instances_newton.usda         # per-mesh physics (Newton)
│   │   ├── materials.usda                # material definitions
│   │   ├── geometries.usd                # mesh data (the big file)
│   │   └── PhysX/ MuJoCo/ Newton/ Physics/   # engine payload subdirs
│   ├── source_assets/                    # original DCC files
│   └── Textures/
└── .thumbs/256x256/                     # rendered thumbnails (lit/unlit)

Tip: for metadata and physics you only need the small text .usda payload files; geometries.usd holds the mesh data and dominates download size.

Naming conventions

Asset names follow <category>_<type>_<variant> (e.g. food_produce_banana_m01, appliance_dishwasher_a01). Variant suffixes (a01, b01, m01, ...) distinguish design variants of the same object type. See NVIDIA's naming guide in the simready-foundation repo for the authoritative conventions.

What's in the catalog

Category (taxonomy group)	Assets
common_assets	2,314
props_isaac	92
props_food	91
omniverse	85
props_warehouse	32
SIM_READY	32
props_kitchenware	6
props_automotive_workshop_tools	5
props_softbody	4
props_endeavor_kitchen	3
props_automotive_parts	2
props_automotive_equipment	1
props_kitchen_france	1
props_hospital	1

Physics

Per-mesh physics properties (physics:mass, physics:centerOfMass, physics:diagonalInertia, collision physics:approximation) are authored in the instances_*.usda payloads. Asset mass below is the sum of per-mesh masses from the PhysX instance file.

Mass (per-asset sum)	Value
Minimum	0.000109254 kg
Median	0.923211 kg
Mean	1790.49 kg
Maximum	952461 kg

Known data quality note: 22 assets (0.82%) report a summed mass above 5000 kg (e.g. warehouse structural modules, an exhaust manifold at ~500,000 kg) — these look like authoring/export defects in NVIDIA's source USD rather than real weights. We report the numbers as-is, unmodified; use the median above rather than the mean when reasoning about a typical asset.

Collision approximations in use: convexDecomposition (9), sdf (10,658)

Heaviest assets

Asset	Mass (kg)
modular_warehouse_h10m_corner_out_a01	952461
exhaustSystem_exhaust_manifold_f01	501226
modular_warehouse_h10m_straight_90_a04	404025
warehouse_h10m_straight_90_a04	400081
warehouse_h10m_straight_a04	399748
modular_warehouse_h10m_straight_a04	398601
modular_warehouse_h10m_center_a01	302944
set_office_main_a01	267161
structure_wall_corner_a01	217653
structure_wall_entrance_a01	192774

Lightest assets

Asset	Mass (kg)
hardware_bearing_caster_a01	0.000109254
factory_nut_m4_loose	0.000267644
factory_nut_m4_tight	0.000268642
hardware_screw_95_a01	0.000406182
hardware_screw_99_physics_a01	0.0005
factory_bolt_m04_loose	0.00063011
screw_high_res	0.000635557
factory_bolt_m04_tight	0.000651707
hardware_bearingPin_a01	0.000784414
factory_nut_m8_loose	0.000992802

Physics engine variants

Each asset declares MuJoCo / Newton / PhysX physics variants in base.usda (Variants → Physics), so the same asset can be simulated in any of the three engines:

Validation (SimReady Foundation)

Assets embed their own validation results (SimReady_Metadata → validation), produced by NVIDIA's simready-foundation tooling against the Robotics-Prop profile. Overall pass rate across parsed validation rows: 55.6%.

Feature	Pass rate	Passed / total
FET000_CORE	91.8%	3,187 / 3,471
FET001_BASE_NEUTRAL	1.5%	54 / 3,690
FET003_BASE_NEUTRAL	99.9%	3,684 / 3,689
FET003_BASE_PHYSX	99.3%	3,610 / 3,637
FET004_BASE_NEUTRAL	42.6%	1,571 / 3,690
FET004_BASE_PHYSX	42.3%	1,538 / 3,637
FET005_BASE_NEUTRAL	65.4%	2,412 / 3,690
FET006_BASE_MDL	1.2%	40 / 3,470
FET100_BASE_ISAACSIM	0.0%	0 / 1

Most common failing requirements:

Failing requirement	Occurrences
RB.MB.001	3,973
AA.001	3,464
VM.MDL.001	3,417
GSP.001	1,190
NP.008	169
HI.009	83
COL.001	38
VM.MAT.001	25
RB.001	20
NP.005	9
VM.TEX.002	3
PMT.001	3
NP.006	2
JT.002	2
ISA.001	1

Growth

Packaging dates extracted from package IDs run 2026-10-08 → 2026-11-02:

How to use

Isaac Sim / Omniverse: add the dataset as a local or cloud asset root; sm_*.usd files are drop-in placeable with physics already tuned.

Plain OpenUSD (Python):

# pip install usd-core
from pxr import Usd, UsdPhysics
stage = Usd.Stage.Open("food_produce_banana_m01/simready_usd/sm_food_produce_banana_m01_01.usd")
for prim in stage.Traverse():
    mass = UsdPhysics.MassAPI(prim)
    if mass:
        print(prim.GetPath(), mass.GetMassAttr().Get())

MuJoCo / Newton: load the same sm_*.usd and activate the MuJoCo (or Newton) physics variant; or use payloads/instances_mujoco.usda directly.

License & attribution

The assets are © NVIDIA Corporation, licensed CC-BY-4.0. When using assets (or derived data such as this catalog), credit "NVIDIA SimReady Assets" and link to https://huggingface.co/datasets/nvidia/simready-assets.

Citation

@dataset{nvidia_simready_assets,
  title  = {NVIDIA SimReady Assets},
  author = {NVIDIA Corporation},
  year   = {2026},
  url    = {https://huggingface.co/datasets/nvidia/simready-assets},
  license = {CC-BY-4.0}
}

Companion resources

Catalog parquets (per-asset physics/validation/size tables): see this project's companion dataset
Interactive stats portal: see this project's Space
SimReady Foundation (validation specs & tooling): https://github.com/NVIDIA/simready-foundation
Isaac Sim: https://developer.nvidia.com/isaac/sim

Generated from repository data as of 2026-11-04; the dataset grows daily — re-run the pipeline for fresh numbers.