THULab/makermods_bimanual_test_booth

Bimanual Test Booth TsFile

This repository presents the numeric time-series portion of makermods/bimanual_test_booth, a LeRobot v3.0 bimanual SO follower robot dataset. Sixteen episodes demonstrate putting each sock into the right bin; one short source episode carries the task label ihjoho.

Source And Attribution

• Original dataset: makermods/bimanual_test_booth
• Repository namespace and publisher: makermods
• The source card names no separate individual author.
• License: Apache-2.0
• Tasks: put each sock into the right bin (16 episodes, 11,476 frames) and ihjoho (1 episode, 634 frames)
• Split: train; 17 episodes, 12,110 frames, 2 task indexes, sampled at 30 fps
• Numeric source: 3 Parquet shards totaling 413,855 bytes
• Source numeric layout: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• The source card provides no paper, homepage, or completed citation.

Data Layout

The data/makermods_bimanual_test_booth.tsfile table is makermods_bimanual_test_booth. It contains 12,110 rows over 17 episode devices. Time is round(timestamp * 1000) in milliseconds, starts at zero for each episode, and increases within each episode. The source timestamp column is omitted because it equals Time / 1000 seconds.

Column	Role	Type	Source mapping
Time	TIME	TIMESTAMP	round(timestamp * 1000) ms
episode_index, task_index	TAG	STRING	Original int64 columns
frame_index	FIELD	INT64	Preserved
sample_index	FIELD	INT64	Renamed from source index
action_0 ... action_11	FIELD	FLOAT	Flattened from action[12]
observation_state_0 ... observation_state_11	FIELD	FLOAT	Flattened from observation.state[12]

The action and state vectors contain the six left-arm joints followed by the same six right-arm joints: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, and gripper.pos.

Source Videos

Videos remain exclusively in the original Hugging Face dataset and are not included here. The source paths are:

• videos/observation.images.left_front/
• videos/observation.images.left_wrist/
• videos/observation.images.left_side/

Each stream is AV1, 640x480, 30 fps, without audio. The source repository has 9 MP4 files totaling 222,396,176 bytes. Video files use videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4. Use episode_index, frame_index, and the original episode metadata to align numeric samples with frames.

Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/makermods_bimanual_test_booth.tsfile")
with reader.query_table(
    "makermods_bimanual_test_booth",
    ["episode_index", "task_index", "frame_index", "sample_index", "action_0"],
    batch_size=65536,
) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()

Citation

The source dataset does not provide a paper or completed citation. Cite the original makermods/bimanual_test_booth dataset and its repository owner, makermods.