THULab/ffw_sg2_rev1_PickCoke2

FFW SG2 Rev1 PickCoke2 TsFile

This dataset is an Apache TsFile conversion of the LeRobot dataset Dongkkka/ffw_sg2_rev1_PickCoke2, published on Hugging Face by Dongkkka.

Modalities: Time-series. The converted repository contains numeric robot observations, actions, frame timing, task/episode tags, and source metadata. Camera videos are not included in the converted repository.

Source Dataset

• Original repository: Dongkkka/ffw_sg2_rev1_PickCoke2
• Original dataset publisher/repository owner: Dongkkka
• License: Apache-2.0
• Robot type: aiworker
• LeRobot codebase version: v2.1
• Task: Pick a coke can and place it in the yellow box.
• Sampling rate: 30 fps
• Available local data: 48 episode Parquet files and 9,767 frames
• Actual episode TAG values: 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 48
• 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
• Video streams described by source metadata: observation.images.cam_head, observation.images.cam_wrist_left, and observation.images.cam_wrist_right

Source Snapshot Consistency

The downloaded snapshot is internally inconsistent. meta/info.json declares 49 episodes and 9,965 frames, while the available Parquet data contains 48 files and 9,767 rows. In addition, episode_000047.parquet contains episode_index=48, while meta/episodes.jsonl describes indexes 0 through 47. This conversion preserves all actual Parquet rows and TAG values without renumbering or fabricating data.

Converted Files

• TsFile: data/ffw_sg2_rev1_pickcoke2.tsfile
• Table: ffw_sg2_rev1_pickcoke2
• Rows: 9,767
• Episodes represented: 48
• Tasks: 1
• Source Parquet files merged: 48
• Time precision: milliseconds
• Metadata: meta/ is mirrored from the source, with meta/info.json updated to describe the converted TsFile and the source inconsistency.

Schema

Time is synthesized as round(timestamp * 1000) milliseconds and restarts from zero in each episode.

TAG columns:

• episode_index
• task_index

Scalar FIELD columns:

• frame_index
• sample_index, renamed from source column index

Flattened FLOAT FIELD groups:

• action[22] -> action_0 ... action_21
• observation.state[22] -> observation_state_0 ... observation_state_21

Conversion Notes

• All 48 available source episode Parquet files are merged into one table-model TsFile. Filter by episode_index and task_index to select an episode or task.
• Vector columns are flattened to scalar TsFile fields. Source column prefixes are preserved, with . replaced by _.
• Source column timestamp is dropped after Time synthesis because it is redundant with Time / 1000 seconds.
• Source column index is renamed to sample_index.
• No available numeric row, state dimension, or action dimension is dropped.
• Camera pixels are not stored in TsFile. Videos are not mirrored in this converted repository; they remain available in the original dataset's videos/ tree. episode_index plus frame_index preserves alignment with the original per-episode videos.

Validation

The converted TsFile was read back with the Apache TsFile Java SDK. Its table schema contains 48 non-time columns (2 TAG and 46 FIELD columns), and query readback matched the staged Parquet at 9,767 rows.

Usage

from tsfile import TsFileReader

path = "data/ffw_sg2_rev1_pickcoke2.tsfile"
reader = TsFileReader(path)

schemas = reader.get_all_table_schemas()
table_name = "ffw_sg2_rev1_pickcoke2"
columns = [
    column.get_column_name()
    for column in schemas[table_name].get_columns()
    if column.get_column_name() != "Time"
]

with reader.query_table(table_name, columns, batch_size=65536) as result:
    batch = result.read_arrow_batch()