taco_play

Converted from the Hugging Face dataset lerobot/taco_play to Apache TsFile format.

Modalities: Time-series.

Source Dataset

The source TACO Play dataset was created using LeRobot. The original dataset page links to the Kaggle homepage at https://www.kaggle.com/datasets/oiermees/taco-robot and the papers https://arxiv.org/abs/2209.08959 and https://arxiv.org/abs/2210.01911.

• License: cc-by-4.0
• Source scale: 3,603 episodes, 237,798 frames, 406 tasks, 15 fps
• Source robot type: unknown
• Source video streams: observation.images.rgb_static and observation.images.rgb_gripper

Converted Files

• data/taco_play.tsfile: one TsFile table containing all train episodes
• meta/info.json: source metadata rewritten to describe the TsFile artifact
• meta/stats.json, meta/tasks.parquet, meta/episodes/chunk-000/file-000.parquet: source metadata mirrored for reference

Videos are not included in this repository. They remain in the original Hugging Face dataset under https://huggingface.co/datasets/lerobot/taco_play/tree/main/videos.

TsFile Schema

• Time: millisecond timestamp, synthesized as round(timestamp * 1000).
• TAG columns: episode_index, task_index.
• FIELD columns: frame_index, sample_index, next_reward, next_done, observation_state_0..observation_state_6, and action_0..action_6.

Vector columns preserve the source column name with . replaced by _, then append the element index. observation.state[7] becomes observation_state_0..observation_state_6; action[7] becomes action_0..action_6. Vector values are stored as single-precision FLOAT.

The source timestamp column is dropped because it is redundant with Time. The source index column is renamed to sample_index. Source video feature columns are omitted from the TsFile table and remain available in the original dataset.

Read Example

Use the Apache TsFile Java or Python SDK to read:

data/taco_play.tsfile

The table name is taco_play. Query a single episode by filtering the TAG:

WHERE episode_index = 0

Citation

@inproceedings{rosete2022tacorl,
    author = {Erick Rosete-Beas and Oier Mees and Gabriel Kalweit and Joschka Boedecker and Wolfram Burgard},
    title = {Latent Plans for Task Agnostic Offline Reinforcement Learning},
    journal = {Proceedings of the 6th Conference on Robot Learning (CoRL)},
    year = {2022}
}
@inproceedings{mees23hulc2,
    title = {Grounding Language with Visual Affordances over Unstructured Data},
    author = {Oier Mees and Jessica Borja-Diaz and Wolfram Burgard},
    booktitle = {Proceedings of the IEEE International Conference on Robotics and Automation (ICRA)},
    year = {2023},
    address = {London, UK}
}