pickup_carrot_remove_parquet_metadata_2

Pickup Carrot (Remove Parquet Metadata 2) (TsFile)

Apache TsFile version of argus-systems/pickup-carrot-remove-parquet-metadata-2.

Overview

A Trossen AI stationary-arm dataset for a single pick-and-place task: pick up a carrot and put it in the bowl. Each frame records the 14-D dual-arm action and 14-D observation state.

• Scale: 58 converted episodes, 25,922 frames, 1 task, sampled at 30 fps; one train split in data/pickup_carrot_remove_parquet_metadata_2.tsfile.

Task (task_index 0): "Pick up carrot and put it in the bowl". Note: the source meta/info.json and meta/episodes.jsonl declare 61 episodes / 27,261 frames, but the source repository only contains 58 data/chunk-000/episode_*.parquet files (episodes 0, 39 and 50 are absent); the 58 present episodes (25,922 frames) were converted.

Schema (TsFile structure)

• Time (INT64, milliseconds) - per-episode sample time, round(timestamp * 1000) in milliseconds (the source timestamp column equals Time / 1000 seconds and is therefore not stored separately).
• episode_index, task_index (TAG) - device dimensions identifying one demonstration episode and one task; query a single episode with WHERE episode_index=0.
• action_0..13, observation_state_0..13 (FIELD) - FLOAT 14-D action and 14-D observation state in the Trossen dual-arm joint space.
• frame_index, sample_index (INT64, FIELD) - source frame counter and source row counter (index renamed to sample_index).

The dataset is stored as one wide TsFile table; within the train split all episodes share that single file with episode_index and task_index as TAG columns. State/action vectors are flattened to one FLOAT field per element (. -> _, element index appended). Camera video files are not included in this repository; the original dataset's videos are at videos/. meta/ is mirrored from the source. No data rows or non-video columns are dropped, apart from the redundant source timestamp column (replaced by Time). Four camera image/video features (observation.images.*) are omitted as non-time-series.

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/pickup_carrot_remove_parquet_metadata_2.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/argus-systems/pickup-carrot-remove-parquet-metadata-2
• Author / publisher: argus-systems
• License: apache-2.0