THULab/galbot_pick_place_demos

Galbot Pick and Place Demos (TsFile)

Apache TsFile version of yizhouzhao-nv/galbot-pick-place-demos.

Overview

A multi-task LeRobot-style manipulation dataset of Galbot robot demonstrations. The repository nests 11 task sessions under lerobot/<task>/, each with its own data/, meta/, and videos/ tree. Every frame holds the commanded action, the observed observation.state, and the observation.environment_state, plus three synchronized camera views (left wrist, right wrist, stage) that are stored as MP4 videos in the original dataset.

The 11 tasks: bar_pull_galbot, pick_place_bottle_bin_galbot, pick_place_can_box_galbot, pick_place_cup_box_galbot, pick_place_cupboard_store_galbot, pick_place_fruit_sort_galbot, pick_place_grape_box_galbot, pick_place_sort_cubes_galbot, pick_place_stack_blocks_galbot, pick_place_stack_cups_galbot, pick_place_weigh_fruit_galbot.

• Tasks: 11 (one per TsFile)
• Episodes: 100
• Frames: 47,930
• Sampling rate: 20 fps

Schema (TsFile structure)

Each task session is converted to one TsFile so episode_index values that restart across tasks remain unambiguous by file/table. Within a TsFile, episode_index and task_index are TAG columns; query a single episode with WHERE episode_index = N.

• Time (INT64, milliseconds) — round(timestamp * 1000); the source timestamp column is dropped (it equals Time / 1000).
• episode_index (TAG) — episode identifier.
• task_index (TAG) — task identifier (single task per session).
• frame_index (INT64) — per-episode frame counter.
• sample_index (INT64) — source index column, renamed.
• action_0..action_15 (FLOAT) — commanded joint positions (16-dim).
• observation_state_0..observation_state_{N-1} (FLOAT) — observed state; the dimension N varies by task (63, 70, or 77).
• observation_environment_state_0..observation_environment_state_20 (FLOAT) — environment state (21-dim).

Dots in source feature names are replaced by underscores and the element index is appended; vector values are stored as single-precision FLOAT.

Usage

Install the Apache TsFile Python SDK (pip install tsfile) and read a converted file:

from pathlib import Path
from tsfile import TsFileReader

path = Path("data/galbot_pick_place_demos_bar_pull_galbot.tsfile")
with TsFileReader(str(path)) as reader:
    schemas = reader.get_all_table_schemas()
    print("tables:", list(schemas))
    table_name = next(iter(schemas))
    table = schemas[table_name]
    columns = [column.get_column_name() for column in table.get_columns()]
    print("columns:", columns)
    field_names = [
        column.get_column_name()
        for column in table.get_columns()
        if column.get_column_name() not in {"Time", "time"}
    ]
    if field_names:
        with reader.query_table(table_name, field_names[:3], batch_size=1024) as result:
            batch = result.read_arrow_batch()
            if batch is not None:
                print(batch.to_pandas().head())

Source & license

• Original dataset: https://huggingface.co/datasets/yizhouzhao-nv/galbot-pick-place-demos
• Author / publisher: yizhouzhao-nv
• License: not declared by the original dataset; please defer to the original.
• Note: camera videos (left wrist / right wrist / stage) are NOT included; see the original dataset's lerobot/<task>/videos/ directories.