The dataset viewer is not available for this subset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
~~~~~~~~~~~~~~~~~~~~~~~~~^
StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 80, in _split_generators
raise ValueError(
...<2 lines>...
)
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
~~~~~~~~~~~~~~~~~~~~~~~^
path=dataset,
^^^^^^^^^^^^^
config_name=config,
^^^^^^^^^^^^^^^^^^^
token=hf_token,
^^^^^^^^^^^^^^^
)
^
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
path,
...<6 lines>...
**config_kwargs,
)
File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.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.
pickbananaV2 — 197 selected raw Piper demonstrations
Task: grasp the banana from the table and place it on the plate.
Upload in progress. Check upload_status.json; use the final manifest.json only after completion.
This release contains 197 original single-arm, two-camera recordings, after excluding three episodes. It contains the original PNG and JSON records, not the experimental 15 Hz aligned derivative, not synthetic segments, and not a ready-to-train LeRobot/RLDX dataset.
Each epN.tar extracts to epN/color/{head,hand}, epN/depth/{head,hand},
epN/joint, and epN/end_pose. Main view is head; wrist view is hand.
The retained source inventory is 232,364 files, totaling 22,655,509,567 bytes before TAR overhead.
Sampling and units
- RGB: uint8 PNG, 640x480, approximately 15 FPS per camera.
- Depth: uint16 PNG, 640x480. Invalid depth pixels are present; the wrist view can have substantial invalid depth.
- Joint feedback: approximately 30 Hz. JSON
positioncontains six joint angles in 0.001 degrees, followed by gripper opening in 0.001 mm. Negative gripper readings are retained as recorded; no zero-offset correction or normalization has been applied. - End pose: original
X,Y,Z,roll,pitch,yawJSON values are unchanged. - There is no separately recorded commanded-action stream. Treating next-step measured positions as training actions is a downstream label-construction choice.
Time synchronization limitations
Filenames contain host-side recording timestamps, not verified common-clock hardware capture times. Original camera exposure times and original SDK feedback timestamps were not preserved in these files. Image/state delays are therefore uncertain. Temporal matching or interpolation cannot prove strict absolute-time synchronization. Image and robot sample counts differ; do not align by file index.
Exclusions and quality checks
- ep6: missing wrist-camera streams; excluded previously.
- ep121: about 2.84 seconds, sampled start/end images show the banana remaining on the table; excluded as an incomplete demonstration.
- ep150: an internal approximately 3.37-second joint/end-pose sampling gap; excluded.
The remaining 197 episodes preserve their original episode boundaries. Episodes with small timing irregularities have not been deleted wholesale. Some startup feedback discontinuities remain, including in ep12 and ep160, and require handling during training conversion.
A local full-file audit reported no image decoding or numeric-format errors for the retained inventory. This is a format/integrity check, not proof of sensor synchronization, successful completion of every demonstration, or validated training quality. Only a subset of episodes was visually inspected.
No additional license is granted by this upload. Final manifest.json records per-archive SHA-256,
size, and original file count. Remote archives are verified before the upload is marked complete.
- Downloads last month
- 112