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
task: string
seed: int64
env_steps: int64
complete_episodes: int64
step_share: struct<0: double, 1: double, 2: double, 3: double, 4: double, 5: double>
  child 0, 0: double
  child 1, 1: double
  child 2, 2: double
  child 3, 3: double
  child 4, 4: double
  child 5, 5: double
by_driver: struct<teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_ (... 762 chars omitted)
  child 0, teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
      child 0, episodes: int64
      child 1, success: double
      child 2, finished: double
      child 3, overflowed: double
      child 4, over_spilled: double
      child 5, error: double
      child 6, spilled: double
  child 1, sensor_teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
      child 0, episodes: int64
      child 1, success: double
      child 2, finished: double
      child 3, overflowed: double
      child 4, over_spilled: double
      child 5, error: double
      child 6, spilled: double
  child 2, noise: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
      child 0, episodes: int64
      child 1, success: double
      child 2, finished: double
      child 3, overflowed: double
      child 4, over_spilled: double
      child 5, error: double
      ch
...
child 0, item: int64
      child 14, decal_area: list<item: list<item: double>>
          child 0, item: list<item: double>
              child 0, item: double
      child 15, decal_keepout: list<item: null>
          child 0, item: null
      child 16, props: list<item: int64>
          child 0, item: int64
      child 17, prop_side: list<item: double>
          child 0, item: double
      child 18, key_power: list<item: double>
          child 0, item: double
      child 19, fill_power: list<item: double>
          child 0, item: double
      child 20, ambient: list<item: double>
          child 0, item: double
      child 21, light_tint: double
      child 22, key_pos: list<item: double>
          child 0, item: double
      child 23, key_jitter: list<item: double>
          child 0, item: double
      child 24, palettes: struct<robot/cup_shell: list<item: list<item: double>>, bottle_shell: list<item: list<item: double>> (... 1 chars omitted)
          child 0, robot/cup_shell: list<item: list<item: double>>
              child 0, item: list<item: double>
                  child 0, item: double
          child 1, bottle_shell: list<item: list<item: double>>
              child 0, item: list<item: double>
                  child 0, item: double
      child 25, palette_jitter: double
      child 26, exposure: list<item: double>
          child 0, item: double
      child 27, white_balance: double
      child 28, image_noise: list<item: double>
          child 0, item: double
to
{'variant': Value('string'), 'env': Value('string'), 'robot': Value('string'), 'control_hz': Value('float64'), 'instructions': List(Value('string')), 'drivers': {'0': Value('string'), '1': Value('string'), '2': Value('string'), '3': Value('string'), '4': Value('string'), '5': Value('string')}, 'terms': {'joint_pos': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'joint_vel': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'ee': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'grip': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'last_action': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'line': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': 
...
 {'camera_period': Value('int64'), 'views': List(List(Value('float64'))), 'look_at': List(Value('float64')), 'view_jitter': Value('float64'), 'look_jitter': Value('float64'), 'roll_jitter': Value('float64'), 'fovy': List(Value('float64')), 'texture_repeat': List(Value('float64')), 'tint': Value('float64'), 'albedo': List(Value('float64')), 'table_half': List(List(Value('float64'))), 'table_center': List(Value('float64')), 'table_shift': Value('float64'), 'decals': List(Value('int64')), 'decal_area': List(List(Value('float64'))), 'decal_keepout': List(Value('null')), 'props': List(Value('int64')), 'prop_side': List(Value('float64')), 'key_power': List(Value('float64')), 'fill_power': List(Value('float64')), 'ambient': List(Value('float64')), 'light_tint': Value('float64'), 'key_pos': List(Value('float64')), 'key_jitter': List(Value('float64')), 'palettes': {'robot/cup_shell': List(List(Value('float64'))), 'bottle_shell': List(List(Value('float64')))}, 'palette_jitter': Value('float64'), 'exposure': List(Value('float64')), 'white_balance': Value('float64'), 'image_noise': List(Value('float64'))}}, 'code_version': Value('string'), 'seed': Value('int64'), 'envs': Value('int64'), 'left_out_as_derived': List(Value('string')), 'clipped_to_card_range': {'wrist_raw': List(Value('float64')), 'torque_raw': List(Value('float64'))}, 'hidden_params': List(Value('string')), 'realism': Value('bool'), 'code_sha256': Value('string'), 'code_snapshot': Value('string'), 'frames': Value('string')}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 149, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                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 129, 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 544, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 3052, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2559, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2584, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 558, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 440, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 361, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2381, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2309, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              task: string
              seed: int64
              env_steps: int64
              complete_episodes: int64
              step_share: struct<0: double, 1: double, 2: double, 3: double, 4: double, 5: double>
                child 0, 0: double
                child 1, 1: double
                child 2, 2: double
                child 3, 3: double
                child 4, 4: double
                child 5, 5: double
              by_driver: struct<teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_ (... 762 chars omitted)
                child 0, teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
                    child 0, episodes: int64
                    child 1, success: double
                    child 2, finished: double
                    child 3, overflowed: double
                    child 4, over_spilled: double
                    child 5, error: double
                    child 6, spilled: double
                child 1, sensor_teacher: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
                    child 0, episodes: int64
                    child 1, success: double
                    child 2, finished: double
                    child 3, overflowed: double
                    child 4, over_spilled: double
                    child 5, error: double
                    child 6, spilled: double
                child 2, noise: struct<episodes: int64, success: double, finished: double, overflowed: double, over_spilled: double, (... 32 chars omitted)
                    child 0, episodes: int64
                    child 1, success: double
                    child 2, finished: double
                    child 3, overflowed: double
                    child 4, over_spilled: double
                    child 5, error: double
                    ch
              ...
              child 0, item: int64
                    child 14, decal_area: list<item: list<item: double>>
                        child 0, item: list<item: double>
                            child 0, item: double
                    child 15, decal_keepout: list<item: null>
                        child 0, item: null
                    child 16, props: list<item: int64>
                        child 0, item: int64
                    child 17, prop_side: list<item: double>
                        child 0, item: double
                    child 18, key_power: list<item: double>
                        child 0, item: double
                    child 19, fill_power: list<item: double>
                        child 0, item: double
                    child 20, ambient: list<item: double>
                        child 0, item: double
                    child 21, light_tint: double
                    child 22, key_pos: list<item: double>
                        child 0, item: double
                    child 23, key_jitter: list<item: double>
                        child 0, item: double
                    child 24, palettes: struct<robot/cup_shell: list<item: list<item: double>>, bottle_shell: list<item: list<item: double>> (... 1 chars omitted)
                        child 0, robot/cup_shell: list<item: list<item: double>>
                            child 0, item: list<item: double>
                                child 0, item: double
                        child 1, bottle_shell: list<item: list<item: double>>
                            child 0, item: list<item: double>
                                child 0, item: double
                    child 25, palette_jitter: double
                    child 26, exposure: list<item: double>
                        child 0, item: double
                    child 27, white_balance: double
                    child 28, image_noise: list<item: double>
                        child 0, item: double
              to
              {'variant': Value('string'), 'env': Value('string'), 'robot': Value('string'), 'control_hz': Value('float64'), 'instructions': List(Value('string')), 'drivers': {'0': Value('string'), '1': Value('string'), '2': Value('string'), '3': Value('string'), '4': Value('string'), '5': Value('string')}, 'terms': {'joint_pos': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'joint_vel': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'ee': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'grip': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'last_action': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': Value('int64'), 'channels': List(Value('string')), 'units': Value('string'), 'derived': Value('bool')}, 'line': {'kind': Value('string'), 'family': Value('string'), 'rate_hz': Value('float64'), 'samples_per_step': 
              ...
               {'camera_period': Value('int64'), 'views': List(List(Value('float64'))), 'look_at': List(Value('float64')), 'view_jitter': Value('float64'), 'look_jitter': Value('float64'), 'roll_jitter': Value('float64'), 'fovy': List(Value('float64')), 'texture_repeat': List(Value('float64')), 'tint': Value('float64'), 'albedo': List(Value('float64')), 'table_half': List(List(Value('float64'))), 'table_center': List(Value('float64')), 'table_shift': Value('float64'), 'decals': List(Value('int64')), 'decal_area': List(List(Value('float64'))), 'decal_keepout': List(Value('null')), 'props': List(Value('int64')), 'prop_side': List(Value('float64')), 'key_power': List(Value('float64')), 'fill_power': List(Value('float64')), 'ambient': List(Value('float64')), 'light_tint': Value('float64'), 'key_pos': List(Value('float64')), 'key_jitter': List(Value('float64')), 'palettes': {'robot/cup_shell': List(List(Value('float64'))), 'bottle_shell': List(List(Value('float64')))}, 'palette_jitter': Value('float64'), 'exposure': List(Value('float64')), 'white_balance': Value('float64'), 'image_noise': List(Value('float64'))}}, 'code_version': Value('string'), 'seed': Value('int64'), 'envs': Value('int64'), 'left_out_as_derived': List(Value('string')), 'clipped_to_card_range': {'wrist_raw': List(Value('float64')), 'torque_raw': List(Value('float64'))}, 'hidden_params': List(Value('string')), 'realism': Value('bool'), 'code_sha256': Value('string'), 'code_snapshot': Value('string'), 'frames': 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.

Sensorcade v0.1 dataset

Rollouts from the Sensorcade simulation benchmark, where robot policies must read many kinds of sensors, live, across five robot bodies. Each variant folder records the raw sensor streams, actions and language instructions of one task variant, as the benchmark's Deploy suite exposes them.

Runs 314, one per variant
Seed folders 328 (314 train, 14 validation)
Size 1.15 TB, 5,886 files
Task families 8, plus 1 long-horizon task
Robots I2RT YAM arm, Unitree G1 humanoid, Unitree Go1 quadruped, Skydio X2 drone, 8-thruster underwater ROV
Physical sensor types 24
License CC-BY-4.0. Two recorded-audio sources are mixed into the sensors; see Attribution.
Code github.com/bosung/sensorcade
Baselines 24 behavior-cloning models trained on this dataset: ultarasonic-penguin/sensorcade-v0.1

Families

Family Robot Runs Seed folders Size
Delicate Grasp YAM arm 85 86 268 GB
Precision Pour YAM arm 45 46 142 GB
Kitchen Prep (long-horizon) YAM arm 1 2 32 GB
Load Carry Unitree G1 37 39 184 GB
Carry and Pour Unitree G1 5 7 40 GB
Disturbed Hover Skydio X2 39 41 123 GB
Gate Racing Skydio X2 1 2 18 GB
Person Following Skydio X2, Unitree Go1 41 44 246 GB
Subsea Inspection 8-thruster ROV 60 61 103 GB

What is recorded

  • Suite. Every variant uses its variant's Deploy suite. Inputs are the robot's own sensors and camera frames (uint8, 128×128, stored on camera ticks). There are no simulator-truth inputs beyond robot state.
  • Sensor realism. It is on for every variant. It adds recorded background sound in the microphones and hydrophone, pitch wobble in tones, rotor vibration in the drone's motor current, and other measured sensor imperfections.
  • Drivers. A teacher drives most steps: a scripted expert, or an RL teacher with privileged state. Perturbed drivers take the remaining steps, so recoveries and failures also appear. The perturbations are action noise, pushes, a weakened teacher and random actions. meta.json names each family's drivers (drivers) and their step shares (driver_shares).
  • Language. Each task has instruction paraphrases (instructions) and the instructions used at evaluation (eval_instructions).
  • Hidden parameters. The physical parameters sampled for each episode are stored for analysis: hidden in the shards, listed in meta.json. They are not policy inputs.

Layout

<Run>/seed<N>/meta.json        Run metadata: per-term sensor kind, true rate, samples per step, channels, units
<Run>/seed<N>/shard_00000.pt   Time-first tensors (torch.save), one shard per block of control steps
<Run>/seed<N>/DONE             Marks a complete recording
variants.json                  Run folder -> released task id, family, robot, seeds, validation seeds
splits.json                    "<Run>/seed<N>" -> "train" or "validation"
SHA256SUMS                     Checksums of every file

Each shard holds:

  • terms: one tensor per observation term
  • action, done, driver, episode_id, step_in_episode and reading_age
  • frame_step and valid, for camera frames
  • hidden

Variant folders and the ids in meta.json use the research-time names (Multisensing-<Env>-..., e.g. folder Box-Lift-Deploy). variants.json gives each variant's released id (e.g. Sensorcade-Load-Carry-Deploy), and sensorcade.ids.new_id maps either form to the released id.

Splits

Seed 3000 of every variant is training data. 14 base variants (each family's base variant and a few others) also have seed 3001, which is validation. splits.json lists the split of every seed folder, and sensorcade.sensor_data.split gives the same split from meta.json. Evaluation uses the benchmark's own seeds and conditions; see the code repository.

Loading

The dataset is 1.15 TB. Fetch all of it, or only the variants you need (variants.json lists them):

hf download ultarasonic-penguin/sensorcade-v0.1 --repo-type dataset --local-dir sensorcade-v0.1 \
  --include "Egg-Pick-Deploy/*" "variants.json" "splits.json" "SHA256SUMS"

Then load it with the benchmark's code:

git clone https://github.com/bosung/sensorcade.git && cd sensorcade && uv sync
export SENSORCADE_DATA=/path/to/sensorcade-v0.1
import json
from sensorcade import sensor_data as sd

variants = sd.scan()  # Every complete seed folder under SENSORCADE_DATA.
train, val, _ = sd.split(variants)

folder = sd.ROOT / "Egg-Pick-Deploy" / "seed3000"
meta = json.loads((folder / "meta.json").read_text())
shard = sd.load(folder / "shard_00000.pt")  # Memory-mapped.
for name, term in sd.parse_terms(meta, shard).items():
  print(name, term.kind, term.rate_hz, tuple(shard["terms"][name].shape))

Sensors

Sensor Card kind Rate (Hz; audio: sample rate) Task families
CTD (conductivity, temperature, depth) ctd 2 Subsea Inspection
Doppler velocity log (DVL) dvl 5 Subsea Inspection
Echo-sounder altimeter echo_altimeter 5 Subsea Inspection
Fingertip tactile pressure array fingertip_pressure 500 Delicate Grasp, Kitchen Prep
Foot pressure insole foot_pressure 10, 500 Load Carry, Carry and Pour
Hydrophone hydrophone 8000 Subsea Inspection
Imaging sonar imaging_sonar 5 Subsea Inspection
IMU imu 50, 200, 500 Load Carry, Carry and Pour, Disturbed Hover, Gate Racing, Person Following, Subsea Inspection
Inertial navigation system (INS) ins_nav 20 Subsea Inspection
Joint encoders joint_encoder 50, 100 Delicate Grasp, Precision Pour, Load Carry, Carry and Pour, Person Following, Kitchen Prep
Joint torque / load joint_torque 50, 500 Delicate Grasp, Precision Pour, Load Carry, Carry and Pour, Kitchen Prep
Lidar lidar 20, 100 Gate Racing, Person Following
Magnetometer magnetometer 20 Subsea Inspection
Microphone (raw audio) microphone 24000 Delicate Grasp, Precision Pour, Kitchen Prep
Motor current motor_current 200, 500 Disturbed Hover, Gate Racing, Person Following, Subsea Inspection
Motor RPM motor_rpm 20 Subsea Inspection
Pressure depth gauge pressure_depth 20 Subsea Inspection
RGB camera rgb_camera 5, 10, 25, 33.3, 100 Delicate Grasp, Precision Pour, Load Carry, Carry and Pour, Disturbed Hover, Gate Racing, Person Following, Subsea Inspection, Kitchen Prep
Scanning sonar scanning_sonar 20 Subsea Inspection
Tether tension tether_tension 20 Subsea Inspection
Thermal camera thermal_camera 9.09, 10 Person Following
USBL acoustic position fix usbl_fix 0.5 Subsea Inspection
Water speed log water_speed_log 10 Subsea Inspection
Wrist force/torque wrist_force 50, 500 Delicate Grasp, Precision Pour, Load Carry, Carry and Pour, Kitchen Prep

Attribution

The dataset is released under CC-BY-4.0. With sensor realism on, sound from two recorded-audio sources is mixed into the simulated sound. Both sources are CC-BY-4.0:

  • Arm microphones (Delicate Grasp, Precision Pour, Kitchen Prep). Background clips cut from REASSEMBLE v1.0.0: Sliwowski et al., Robotics: Science and Systems (RSS) 2025, arXiv:2502.05086, https://researchdata.tuwien.ac.at/records/0ewrv-8cb44. Changes: decoded, cut to 1.5 s looping clips, scaled and mixed with the simulated sound.
  • ROV hydrophone (Subsea Inspection). Ambient clips cut from the MBARI Pacific Ocean Sound Recordings (MARS cabled observatory), courtesy of the Monterey Bay Aquarium Research Institute (MBARI). Citation requested by MBARI: https://doi.org/10.1109/OCEANS.2016.7761363. Changes: decimated to 8 kHz, cut to 10 s looping clips, scaled and mixed with the simulated sound.

The code repository holds the clips, their full attribution and the scripts that cut them, under src/sensorcade/assets/.

Citation

TBA.

Downloads last month
77

Models trained or fine-tuned on ultarasonic-penguin/sensorcade-v0.1

Paper for ultarasonic-penguin/sensorcade-v0.1