The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
frame_id: string
candidates: list<item: struct<T_head_optical_marker: list<item: list<item: double>>, reprojection_rms_px: double (... 100 chars omitted)
child 0, item: struct<T_head_optical_marker: list<item: list<item: double>>, reprojection_rms_px: double, depth_pla (... 88 chars omitted)
child 0, T_head_optical_marker: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 1, reprojection_rms_px: double
child 2, depth_plane_median_abs_error_m: double
child 3, T_left_base_head_optical: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
selected_candidate: int64
selection: string
independent_validation: bool
valid_depth_pixels: int64
head_corner_pixels: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
files: list<item: struct<path: string, bytes: int64, sha256: string>>
child 0, item: struct<path: string, bytes: int64, sha256: string>
child 0, path: string
child 1, bytes: int64
child 2, sha256: string
to
{'files': List({'path': Value('string'), 'bytes': Value('int64'), 'sha256': 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
frame_id: string
candidates: list<item: struct<T_head_optical_marker: list<item: list<item: double>>, reprojection_rms_px: double (... 100 chars omitted)
child 0, item: struct<T_head_optical_marker: list<item: list<item: double>>, reprojection_rms_px: double, depth_pla (... 88 chars omitted)
child 0, T_head_optical_marker: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
child 1, reprojection_rms_px: double
child 2, depth_plane_median_abs_error_m: double
child 3, T_left_base_head_optical: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
selected_candidate: int64
selection: string
independent_validation: bool
valid_depth_pixels: int64
head_corner_pixels: list<item: list<item: double>>
child 0, item: list<item: double>
child 0, item: double
files: list<item: struct<path: string, bytes: int64, sha256: string>>
child 0, item: struct<path: string, bytes: int64, sha256: string>
child 0, path: string
child 1, bytes: int64
child 2, sha256: string
to
{'files': List({'path': Value('string'), 'bytes': Value('int64'), 'sha256': Value('string')})}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Piper Active Vision 100
Incremental release: the target selection contains 100 demonstrations, but not all
100 are uploaded yet. Consult episodes.json and release_status.json for the
published subset. Only rows marked uploaded are complete published episodes.
This initial upload batch contains 100 episodes (166.53 GB of source files).
Tasks and selection
The target selection contains 50 corner demonstrations (C0–C4, ten per layout) and 50 deep-box demonstrations (B0–B4, ten per layout), collected on a Piper robot with the left arm active. Task-local episode numbers run from 001 to 050. Raw storage numbers can differ: the release index preserves that mapping. Failed attempts excluded from this selection and later policy-evaluation trials are not part of this dataset.
- Corner: find the yellow object to the robot's left, grasp and lift it.
- Deep box: find the object inside the box, grasp it, and lift it completely out.
Files and reading
Each data/<task>/episode_NNN/ contains the original source files listed in its
UPLOAD_READY.json, plus a release SHA256SUMS file. Source files are unchanged.
raw.bag is a ROS1 bag containing image sequences and robot topics; this is not
a preconverted LeRobot/Parquet dataset or a set of standalone MP4 videos.
- RGB:
/camera_f/color/image_raw,/camera_l/color/image_raw,/camera_r/color/image_raw(head, left wrist, right wrist). - Head depth:
/camera_f/aligned_depth_to_color/image_raw. - Joint observations:
/puppet/joint_left,/puppet/joint_right. - Command reference:
/master/joint_left. It is not an independently verified motor-applied command; do not substitute observations for actions. left-raw-can.jsonl,right-raw-can.jsonl: raw CAN evidence.episode.json: task, units, operator events, success label and pose association.message-receive-times.jsonl,clock-quality.json: timing evidence.evidence_observation_ready/: images, depth and metadata at the ready event.calibration/: intrinsics/extrinsics and measurement evidence as recorded.
With ROS1 tools, use rosbag info path/to/raw.bag or the Python rosbag.Bag
reader to inspect messages without publishing them to a robot.
For units and per-episode details, read episode.json first. The vendor driver
used an approximate conversion: raw millidegrees / 1000 * 0.017444 for joint
radians; raw CAN permits conversion with pi/180 instead. Gripper conversion is
raw / 1000000 to metres. Do not apply a second conversion to already-scaled fields.
Labels and limitations
Events such as first visibility and observation readiness are operator keyboard annotations, not frame-exact ground truth. Sensor source clocks and receive times can differ; synchronization is required before producing aligned training samples. The two-second hold criterion is operator-confirmed and is not pending a separate frame-by-frame review. This does not claim independent verification of every task outcome or exact event timing.
Auxiliary marker poses are approximate. The frame is the paper marker's X corner, not the object center. Keep the per-layout validation errors and planar-pose ambiguities; absolute pose accuracy was not independently established. Historical head/right calibration and later left-hand-eye candidates may coexist: follow each measurement bundle's declared effective parameters rather than picking a YAML file by name alone.
Corner C3 has partial initial right-wrist target visibility accepted by the collector, and corner episode 046 (C4) has partial initial left-wrist visibility. These are not strict initially-invisible evaluation examples. Layouts remain fixed within each group of ten, so demonstrations are correlated. No train/test split is prescribed; split by layout when assessing layout generalization.
Download and verification
from huggingface_hub import snapshot_download
snapshot_download(
repo_id="renbin1/piper-active-vision-100", repo_type="dataset",
allow_patterns=["README.md", "LICENSE", "episodes.json", "release_status.json",
"data/deepbox/episode_001/**"],
local_dir="piper-active-vision-100",
)
Remove allow_patterns to download all currently published files. Within an
episode directory run sha256sum -c SHA256SUMS to check the files. Upload workers
rehash source files before publishing and verify remote file sizes and available
LFS SHA256 metadata before marking an episode uploaded.
License and attribution
Data is released under Creative Commons Attribution–NonCommercial 4.0
International (CC BY-NC 4.0). Noncommercial sharing and adaptation are allowed
under the license terms; credit renbin1 / Piper Active Vision 100, link this
dataset and the license, and indicate modifications. Commercial use is not
authorized by this license. See LICENSE and
https://creativecommons.org/licenses/by-nc/4.0/ . No paper citation is specified yet.
- Downloads last month
- 2,668