THULab/kivod_so100_test2

SO100 Test2 TsFile

This is the Apache TsFile edition of kivod/so100_test2, a LeRobot v2.1 SO100 dataset for the task "Grasp a cucumber and put it in the hole." Numeric robot trajectories are stored in one table-model TsFile.

Source And Attribution

• Source dataset: kivod/so100_test2
• Author and uploader: kivod
• License: Apache-2.0
• Task: Grasp a cucumber and put it in the hole.
• Split: train
• Scale: 50 episodes, 23,585 frames, 1 task, and 30 FPS
• Source layout: 50 Parquet shards under data/chunk-000/
• TsFile layout: one file, data/kivod_so100_test2.tsfile
• Robot type: so100
• The source card does not provide a homepage, paper, or completed BibTeX citation.

Data Layout

The TsFile contains 23,585 rows in the table kivod_so100_test2. Time is round(timestamp * 1000) in milliseconds and restarts at zero for each episode. The source timestamp column is omitted because it is represented by Time / 1000 seconds. Source index is renamed to sample_index.

Column	Role	Type	Source mapping
Time	TIME	INT64 milliseconds	round(timestamp * 1000)
episode_index, task_index	TAG	STRING	Original INT64 identifiers
frame_index	FIELD	INT64	Preserved frame position
sample_index	FIELD	INT64	Source index
action_0 ... action_5	FIELD	FLOAT	Flattened from action[6]
observation_state_0 ... observation_state_5	FIELD	FLOAT	Flattened from observation.state[6]

Vector prefixes preserve the source names with dots replaced by underscores. The six dimensions remain in this order: main_shoulder_pan, main_shoulder_lift, main_elbow_flex, main_wrist_flex, main_wrist_roll, and main_gripper. No trajectory rows, episodes, tasks, action dimensions, or state dimensions are removed.

Source Videos

Videos are not included here. The source repository contains 50 AV1 MP4 files for observation.images.laptop under videos/chunk-000/observation.images.laptop, from episode_000000.mp4 through episode_000049.mp4. The source template is videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4. Use episode_index and frame_index to align numeric rows with the corresponding 30 FPS video frames.

Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/kivod_so100_test2.tsfile")
with reader.query_table(
    "kivod_so100_test2",
    [
        "episode_index",
        "task_index",
        "frame_index",
        "sample_index",
        "action_0",
        "observation_state_0",
    ],
    batch_size=1024,
) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()