Dataset Viewer
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
gripper_id: int64
left_finger_tag_id: int64
right_finger_tag_id: int64
max_width: double
min_width: double
image_width: int64
image_height: int64
nr_calib_images: int64
intrinsic_type: string
stabilized: bool
fps: int64
intrinsics: struct<aspect_ratio: double, focal_length: double, principal_pt_x: double, principal_pt_y: double, r (... 128 chars omitted)
child 0, aspect_ratio: double
child 1, focal_length: double
child 2, principal_pt_x: double
child 3, principal_pt_y: double
child 4, radial_distortion_1: double
child 5, radial_distortion_2: double
child 6, radial_distortion_3: double
child 7, radial_distortion_4: double
child 8, skew: double
final_reproj_error: double
to
{'final_reproj_error': Value('float64'), 'fps': Value('int64'), 'image_height': Value('int64'), 'image_width': Value('int64'), 'intrinsic_type': Value('string'), 'intrinsics': {'aspect_ratio': Value('float64'), 'focal_length': Value('float64'), 'principal_pt_x': Value('float64'), 'principal_pt_y': Value('float64'), 'radial_distortion_1': Value('float64'), 'radial_distortion_2': Value('float64'), 'radial_distortion_3': Value('float64'), 'radial_distortion_4': Value('float64'), 'skew': Value('float64')}, 'nr_calib_images': Value('int64'), 'stabilized': Value('bool')}
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(
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 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 483, 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 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, 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 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/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/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, 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 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
gripper_id: int64
left_finger_tag_id: int64
right_finger_tag_id: int64
max_width: double
min_width: double
image_width: int64
image_height: int64
nr_calib_images: int64
intrinsic_type: string
stabilized: bool
fps: int64
intrinsics: struct<aspect_ratio: double, focal_length: double, principal_pt_x: double, principal_pt_y: double, r (... 128 chars omitted)
child 0, aspect_ratio: double
child 1, focal_length: double
child 2, principal_pt_x: double
child 3, principal_pt_y: double
child 4, radial_distortion_1: double
child 5, radial_distortion_2: double
child 6, radial_distortion_3: double
child 7, radial_distortion_4: double
child 8, skew: double
final_reproj_error: double
to
{'final_reproj_error': Value('float64'), 'fps': Value('int64'), 'image_height': Value('int64'), 'image_width': Value('int64'), 'intrinsic_type': Value('string'), 'intrinsics': {'aspect_ratio': Value('float64'), 'focal_length': Value('float64'), 'principal_pt_x': Value('float64'), 'principal_pt_y': Value('float64'), 'radial_distortion_1': Value('float64'), 'radial_distortion_2': Value('float64'), 'radial_distortion_3': Value('float64'), 'radial_distortion_4': Value('float64'), 'skew': Value('float64')}, 'nr_calib_images': Value('int64'), 'stabilized': Value('bool')}
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.
TRumi validation — 2026-09-29
Processed output from the first hands-on validation of two Trossen TRumi handheld grippers
(UMI-style: GoPro HERO13 Black on Labs firmware 2.10.70, 2.7K 4:3 at 119.88 fps, ArUco finger tags,
ORB-SLAM3 mono-inertial localization). Code, test plan and the full experiment record:
github.com/ewinchell07/trumi-validation (private).
| Folder | What | Episodes | Status |
|---|---|---|---|
session_1_tabletop/ |
Bimanual pick-and-place on a static table, 10 demos | 9 (both grippers) | Usable. Exclude demo 1 (mapping marker left in view) and demo 9 (deliberately fast/jerky) from training. Demo 10 (deliberately off-map start) was dropped by the pipeline. |
session_2_conveyor/ |
Bimanual conveyor pick-and-place, 20 demos at belt speeds 20/40/60/80 (5 each) | 3 (gripper 0 only) | Not usable as bimanual data. SLAM failed on 38 of 40 camera clips. These 3 single-arm episodes are a salvage, kept for debugging. |
Files per session
dataset_mcap/episode_*.mcap: one MCAP per episode, with camera images (/cameras/camera*/image), end-effector pose (robot*/eef/state) and gripper width (robot*/gripper/state). View in Foxglove.dataset_plan.pkl: the TRumi pipeline's episode plan (timestamps, TCP poses, gripper widths, video spans).calibration/: camera intrinsics, ArUco config, SLAM-to-marker transform, per-camera gripper range.labels.csv: demo number → condition/note.demo_metrics.csv: per-camera-clip SLAM tracking metrics (tracked share, pose jumps > 3 cm, smoothed speed, whether it reached the dataset).session_report.txt: the pipeline's pass/fail report.
Cameras: C3534250043778 = gripper 0 (finger tags 0/1), C3535424690582 = gripper 1 (tags 6/7).
Demo numbers follow each camera's clip order. Raw GoPro clips are not included.
Known issues
- Session 1 pose tracks are jumpier than Trossen's reference: 3–12% of frame-to-frame steps exceed 3 cm, against 0.2–0.6% in Trossen's example dataset.
- Session 2: the ORB-SLAM3 build hangs or flickers after relocalizing mid-clip. See the experiment record.
- Downloads last month
- 251