aloha_zipping_bag

This dataset is a TsFile conversion of the LeRobot dataset locht131/aloha-zipping-bag. The source dataset was created using LeRobot and contains an ALOHA robot bag-zipping task.

Modalities: Time-series.

Source Dataset

• Original dataset: locht131/aloha-zipping-bag
• License: MIT
• Codebase version: v2.1
• Robot type: ALOHA
• Split: train 0:99
• Scale: 99 episodes, 40,313 frames, 1 task, 1 chunk
• Sampling rate: 15 fps
• Source frame layout: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
• Source video layout: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4

Task:

Use both hands to center and align the bag, zip it halfway with the left hand, then pass the zipper to the right hand and finish closing the bag.

Converted Data

• TsFile path: data/aloha_zipping_bag.tsfile
• Table name: aloha_zipping_bag
• Rows: 40,313
• TAG columns: episode_index, task_index
• Time precision: milliseconds
• Time mapping: Time = round(timestamp * 1000)
• Mirrored metadata: meta/info.json, meta/tasks.jsonl, meta/episodes.jsonl, meta/episodes_stats.jsonl

meta/info.json has been updated so that data_path points to data/aloha_zipping_bag.tsfile and includes a tsfile_conversion object documenting the source path, converted path, TAG columns, row count, flattened features, and video policy.

Schema

Time is the TsFile timestamp column. The original episode_index and task_index columns are stored as TAG columns, so downstream queries can select episodes or tasks using TsFile device/tag filters.

Scalar FIELD columns:

• frame_index
• sample_index converted from source index

Flattened FLOAT FIELD groups:

• action_0 through action_13 from source action[14]
• observation_state_0 through observation_state_13 from source observation.state[14]
• observation_velocity_0 through observation_velocity_13 from source observation.velocity[14]
• observation_effort_0 through observation_effort_13 from source observation.effort[14]

Conversion Notes

• The source timestamp column is used only to synthesize Time and is not retained as a separate FIELD because it is redundant with Time / 1000 seconds.
• Source index is renamed to sample_index.
• No numeric time-series rows are intentionally dropped.
• Videos are not included in this converted repository. The original dataset contains 297 AV1 MP4 videos across three video features: observation.images.color.high, observation.images.color.wrist_right, and observation.images.color.wrist_left.
• Source videos remain available in the original dataset under videos/.

Usage

from tsfile import TsFileReader

reader = TsFileReader("data/aloha_zipping_bag.tsfile")
tables = reader.get_all_table_schemas()
print(tables.keys())

Read the .tsfile file with the Apache TsFile SDK or compatible tooling.