Dataset Viewer
Duplicate
The dataset viewer is not available for this split.
Cannot load the dataset split (in streaming mode) to extract the first rows.
Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
total: int64
failed: int64
checks: list<item: struct<check: string, pass: bool, detail: string>>
  child 0, item: struct<check: string, pass: bool, detail: string>
      child 0, check: string
      child 1, pass: bool
      child 2, detail: string
joint: struct<type: string, hinge: list<item: double>, axis: list<item: double>, range_rad: list<item: doub (... 226 chars omitted)
  child 0, type: string
  child 1, hinge: list<item: double>
      child 0, item: double
  child 2, axis: list<item: double>
      child 0, item: double
  child 3, range_rad: list<item: double>
      child 0, item: double
  child 4, effort: double
  child 5, velocity: double
  child 6, gravity_torque_max_Nm: double
  child 7, holding_mechanism: string
  child 8, mjcf_dynamics: struct<damping: double, frictionloss: double, stiffness: double, springref: double, armature: double (... 13 chars omitted)
      child 0, damping: double
      child 1, frictionloss: double
      child 2, stiffness: double
      child 3, springref: double
      child 4, armature: double
      child 5, kp: double
scale_source: string
density: double
material: string
inertia_method: struct<lid: string, base: string>
  child 0, lid: string
  child 1, base: string
gate: struct<hulls: int64, seat_shift_mm: double, axis_tune_mm: double, geom_peak_mm: double, closed_pen_m (... 140 chars omitted)
  child 0, hulls: int64
  child 1, seat_shift_mm: double
  child 2, axis_tune_mm: double
  child 3, geom_peak_mm: double
  child 4, closed_pen_mm: double
  child 5, sweep_pen_mm: double
  child 6, reached_deg: int64
  child 7, target_deg: int64
  child 8, stable: bool
  child 9, upright: bool
  child 10, topple_at_deg: null
  child 11, slam_topple: bool
scale: double
mass_kg: struct<base: double, lid: double>
  child 0, base: double
  child 1, lid: double
formats: list<item: string>
  child 0, item: string
to
{'scale': Value('float64'), 'scale_source': Value('string'), 'density': Value('float64'), 'material': Value('string'), 'mass_kg': {'base': Value('float64'), 'lid': Value('float64')}, 'inertia_method': {'lid': Value('string'), 'base': Value('string')}, 'joint': {'type': Value('string'), 'hinge': List(Value('float64')), 'axis': List(Value('float64')), 'range_rad': List(Value('float64')), 'effort': Value('float64'), 'velocity': Value('float64'), 'gravity_torque_max_Nm': Value('float64'), 'holding_mechanism': Value('string'), 'mjcf_dynamics': {'damping': Value('float64'), 'frictionloss': Value('float64'), 'stiffness': Value('float64'), 'springref': Value('float64'), 'armature': Value('float64'), 'kp': Value('float64')}}, 'gate': {'hulls': Value('int64'), 'seat_shift_mm': Value('float64'), 'axis_tune_mm': Value('float64'), 'geom_peak_mm': Value('float64'), 'closed_pen_mm': Value('float64'), 'sweep_pen_mm': Value('float64'), 'reached_deg': Value('int64'), 'target_deg': Value('int64'), 'stable': Value('bool'), 'upright': Value('bool'), 'topple_at_deg': Value('null'), 'slam_topple': Value('bool')}, 'formats': List(Value('string'))}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                         ^^^^^^^^^
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                         ^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/src/services/worker/src/worker/utils.py", line 478, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2815, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2352, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 2377, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 310, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/json/json.py", line 130, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2369, in table_cast
                  return cast_table_to_schema(table, schema)
                         ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.12/site-packages/datasets/table.py", line 2297, in cast_table_to_schema
                  raise CastError(
              datasets.table.CastError: Couldn't cast
              total: int64
              failed: int64
              checks: list<item: struct<check: string, pass: bool, detail: string>>
                child 0, item: struct<check: string, pass: bool, detail: string>
                    child 0, check: string
                    child 1, pass: bool
                    child 2, detail: string
              joint: struct<type: string, hinge: list<item: double>, axis: list<item: double>, range_rad: list<item: doub (... 226 chars omitted)
                child 0, type: string
                child 1, hinge: list<item: double>
                    child 0, item: double
                child 2, axis: list<item: double>
                    child 0, item: double
                child 3, range_rad: list<item: double>
                    child 0, item: double
                child 4, effort: double
                child 5, velocity: double
                child 6, gravity_torque_max_Nm: double
                child 7, holding_mechanism: string
                child 8, mjcf_dynamics: struct<damping: double, frictionloss: double, stiffness: double, springref: double, armature: double (... 13 chars omitted)
                    child 0, damping: double
                    child 1, frictionloss: double
                    child 2, stiffness: double
                    child 3, springref: double
                    child 4, armature: double
                    child 5, kp: double
              scale_source: string
              density: double
              material: string
              inertia_method: struct<lid: string, base: string>
                child 0, lid: string
                child 1, base: string
              gate: struct<hulls: int64, seat_shift_mm: double, axis_tune_mm: double, geom_peak_mm: double, closed_pen_m (... 140 chars omitted)
                child 0, hulls: int64
                child 1, seat_shift_mm: double
                child 2, axis_tune_mm: double
                child 3, geom_peak_mm: double
                child 4, closed_pen_mm: double
                child 5, sweep_pen_mm: double
                child 6, reached_deg: int64
                child 7, target_deg: int64
                child 8, stable: bool
                child 9, upright: bool
                child 10, topple_at_deg: null
                child 11, slam_topple: bool
              scale: double
              mass_kg: struct<base: double, lid: double>
                child 0, base: double
                child 1, lid: double
              formats: list<item: string>
                child 0, item: string
              to
              {'scale': Value('float64'), 'scale_source': Value('string'), 'density': Value('float64'), 'material': Value('string'), 'mass_kg': {'base': Value('float64'), 'lid': Value('float64')}, 'inertia_method': {'lid': Value('string'), 'base': Value('string')}, 'joint': {'type': Value('string'), 'hinge': List(Value('float64')), 'axis': List(Value('float64')), 'range_rad': List(Value('float64')), 'effort': Value('float64'), 'velocity': Value('float64'), 'gravity_torque_max_Nm': Value('float64'), 'holding_mechanism': Value('string'), 'mjcf_dynamics': {'damping': Value('float64'), 'frictionloss': Value('float64'), 'stiffness': Value('float64'), 'springref': Value('float64'), 'armature': Value('float64'), 'kp': Value('float64')}}, 'gate': {'hulls': Value('int64'), 'seat_shift_mm': Value('float64'), 'axis_tune_mm': Value('float64'), 'geom_peak_mm': Value('float64'), 'closed_pen_mm': Value('float64'), 'sweep_pen_mm': Value('float64'), 'reached_deg': Value('int64'), 'target_deg': Value('int64'), 'stable': Value('bool'), 'upright': Value('bool'), 'topple_at_deg': Value('null'), 'slam_topple': Value('bool')}, 'formats': List(Value('string'))}
              because column names don't match

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.

njc-articulated-hinge: sim-verified articulated 3D objects

Fifteen household objects rebuilt as physics-valid articulated assets for robot learning. Each object has one revolute joint (a lid, door, or clamshell hinge) and ships as URDF and native MuJoCo MJCF with textured visual meshes, convex collision decomposition, computed mass and inertia, a tuned joint, and its own actuator. Every object was actuated through its full range in MuJoCo and had to pass an automated 26-check physics gate before inclusion. Executability: 15/15.

Built by notjustchairs.ai.

Why this dataset

Most articulated-object assets are kinematics-only: joints with a type, axis, and limits, but no validated dynamics. Loaded into a physics simulator they interpenetrate, jam, fly apart, or fall open under gravity. This set inverts the priority: the simulator is the gatekeeper. An asset only ships if, loaded cold from the published files with contacts fully active, it:

  • rests closed on the floor without interpenetration (worst-case contact depth budgeted at 5 mm),
  • actuates through at least 90% of its authored joint range,
  • holds position when released with the actuator disabled (no gravity flop),
  • survives a full-force slam without going unstable,
  • does not topple from its own articulation (instability is recorded when present),
  • and passes structural checks on meshes, textures, inertia tensors, and file formats.

The 15 objects

Class Objects
Lid (8) wooden box, jewelry box, toolbox, bread box, cooler, toy chest, trash bin, toilet
Door (6) kitchen cabinet, medicine cabinet, minifridge, microwave, safe, locker
Clamshell (1) laptop

What's in each asset

Each asset is a folder under assets/<id>/:

File Description
object.urdf Free-standing URDF: two links, one revolute joint with limits, effort/velocity bounds, damping and friction, explicit inertials.
object.xml Native MuJoCo MJCF: free-floating base, hinge joint with full dynamics, position actuator, visual and collision geoms.
scene.xml MuJoCo scene wrapper (object + floor + light), Menagerie-style; load this to simulate immediately.
base.glb, lid_closed.glb Textured visual meshes (movable part baked at the closed pose).
base_final.obj, lid_final.obj (+ .mtl, .png) Visual meshes referenced by URDF/MJCF.
collision/*.obj Convex decomposition (CoACD), separate from visuals.
web_joint.json Hinge point, axis, and limit for browser viewers.
meta.json Scale provenance, material, masses, inertia method, joint schema, gravity-torque analysis, holding mechanism, and the full QC gate record.
qc.json The 26-check validation report for this object.
sim.gif The MuJoCo verification sweep, rendered.
source.png The concept image the object was reconstructed from.

A top-level metadata.jsonl holds one row per asset; stats.json has aggregate numbers.

Physics fields worth knowing

  • Real-world scale: every object is metrically scaled from a user-provided real height (scale_source: "user"), not normalized units.
  • Mass and inertia: computed from the actual mesh shell (voxel-shell method for non-watertight parts, recorded per part in inertia_method), optionally rescaled to user-provided masses. A hollow box is not a solid block here.
  • Joint dynamics: each joint records its maximum gravity torque and a holding_mechanism. Gravity-opening joints receive dry friction or a counterbalance spring (in the MJCF; URDF has no native spring element, so URDF consumers should read meta.json and inject equivalents at load time, the same pattern ManiSkill uses for PartNet-Mobility).
  • Collision honesty: convex decomposition cannot represent every nesting shape perfectly; where a simplification was applied it is recorded in meta.json (collision_note), never silent.

Intended uses

  • Robot-learning environments: drop scene.xml into MuJoCo and train open/close manipulation immediately; URDF for other simulators.
  • Articulated-object understanding: paired source image, textured mesh, parts, and joint parameters per object.
  • Scene composition: retrieval-ready articulated assets for simulation scene generators.

Limitations

  • 15 objects, one revolute joint each. No prismatic joints, no multi-joint objects in this release.
  • Collision is convex decomposition: shapes that nest (a flip-lid skirt over a rim) carry a documented clearance floor in meta.json.
  • Visual meshes are single-image 3D reconstructions: stylized, occasional small surface artifacts, not CAD or scans of specific real products.
  • Masses and materials are assigned per object (recorded with provenance), not measured from physical references.
  • Validated in MuJoCo. URDF loads cleanly elsewhere, but other simulators' defaults differ; PhysX/Isaac validation is planned, not included.

License & attribution

Released under CC-BY-4.0:

"njc-articulated-hinge: sim-verified articulated 3D objects" by notjustchairs.ai, CC-BY-4.0.

Reconstruction and processing use open tools and models with permissive licenses, including TRELLIS (MIT outputs), BiRefNet, CoACD, and MuJoCo. See THIRD-PARTY-LICENSES.md.

Citation

@misc{notjustchairs_articulatedhinge_2026,
  title  = {njc-articulated-hinge: sim-verified articulated 3D objects},
  author = {notjustchairs.ai},
  year   = {2026},
  howpublished = {Hugging Face Datasets},
  note   = {CC-BY-4.0}
}

Dataset statistics

  • 15 objects: 8 lids, 6 doors, 1 clamshell
  • 390 automated QC checks passed (26 per object, zero failures)
  • Executability: 15/15 (full-range actuation in MuJoCo from the shipped files, contacts active)
  • 1296 convex collision hulls across the set
  • All objects metrically scaled with user provenance; masses assigned per object
Object Class Mass (base+lid kg) Open range Hulls QC
laptop clamshell 3.0 + 1.5 105 deg 53 26/26
kitchen cabinet door 6.0 + 1.6 105 deg 124 26/26
locker door 25.0 + 4.0 127 deg 49 26/26
medicine cabinet door 8.0 + 2.0 80 deg 71 26/26
microwave door 12.0 + 1.5 83 deg 95 26/26
minifridge door 18.0 + 3.0 113 deg 59 26/26
safe door 35.0 + 7.0 89 deg 82 26/26
bread box lid 1.5 + 0.5 113 deg 75 26/26
cooler lid 4.0 + 1.5 129 deg 121 26/26
jewelry box lid 0.5 + 0.1 142 deg 53 26/26
toilet lid 25.0 + 2.0 94 deg 124 26/26
toolbox lid 3.0 + 0.8 130 deg 60 26/26
toy chest lid 7.0 + 2.0 171 deg 130 26/26
trash bin lid 2.5 + 0.4 120 deg 110 26/26
wooden box lid 3.2 + 1.3 139 deg 90 26/26

Questions, custom articulated collections, or verified data for your pipeline? Get in touch.

Downloads last month
4