THULab/pick_place_cube_1_18

pick_place_cube_1.18 TsFile

Converted from the Hugging Face dataset CSCSXX/pick_place_cube_1.18 to Apache TsFile format.

The source dataset was created with LeRobot. It is an SO-100 robot pick-place-cube dataset with one task: pick-place-cube.

Dataset Facts

• Modalities: Time-series. Source video streams are available only in the original dataset.
• License: Apache-2.0.
• Robot type: so100.
• Codebase version: v2.0.
• Episodes: 50.
• Frames / converted rows: 16,491.
• Tasks: 1.
• Source split: train episodes 0:50.
• Sampling rate: 30 fps.
• Source videos: 100 videos from observation.images.top and observation.images.side.
• Converted file: data/pick_place_cube_1_18.tsfile.

Converted Schema

The converted TsFile table is named pick_place_cube_1_18.

• Time: int64 millisecond timestamp, synthesized as round(timestamp * 1000).
• TAG columns: episode_index, task_index.
• FIELD columns: frame_index, sample_index, action_0 to action_5, and observation_state_0 to observation_state_5.

The original vector columns are flattened as scalar fields:

• action with shape [6] becomes action_0 to action_5.
• observation.state with shape [6] becomes observation_state_0 to observation_state_5.

Conversion Notes

• Numeric LeRobot frame Parquet files from data/chunk-*/*.parquet were merged into one TsFile.
• Time uses millisecond precision and restarts per episode, matching the source timestamp convention.
• The source timestamp column is not retained as a separate field because it is redundant with Time / 1000 seconds.
• The source index column is renamed to sample_index.
• Source meta/ files are mirrored, with meta/info.json updated to describe the converted TsFile artifact and original video location.
• Video streams are not uploaded to this repository. They remain in the original Hugging Face dataset under videos/.

Read Example

from tsfile import TsFileReader

path = "data/pick_place_cube_1_18.tsfile"
with TsFileReader(path) as reader:
    schemas = reader.get_all_table_schemas()
    print(schemas.keys())