jzpeng/max-chr_libero_plus_10_add_object_all

LIBERO Plus 10 Add Object All TsFile

This dataset is an Apache TsFile conversion of max-chr/libero_plus_10_add_object_all, a LeRobot-format Franka manipulation dataset with ten task instructions, 417 episodes, and 112,885 frame rows sampled at 20 fps.

Modalities: Time-series; Tabular. Numeric robot observations, actions, frame timing, episode/task tags, and source metadata are included. Camera videos remain in the original Hugging Face dataset.

Source Dataset and Attribution

• Original dataset: max-chr/libero_plus_10_add_object_all
• Pinned source revision: 7be02532d3c4586cc6fa70c26c92ae6385a64ecf (files and versions)
• Original uploader/author: Maximilian Christof (max-chr)
• License: Apache-2.0
• LeRobot codebase version: v3.0; robot type: franka
• The source card does not provide a completed paper or BibTeX citation.
• Frame template: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• Video template: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

Tasks and Scale

The train split covers 417 episodes (episode_index 0–416), 112,885 rows, and 10 task indexes.

task_index	Instruction	Episodes	Frames
0	put both the cream cheese box and the butter in the basket	52	13,455
1	put both the alphabet soup and the cream cheese box in the basket	50	13,368
2	put both moka pots on the stove	32	13,192
3	put the white mug on the plate and put the chocolate pudding to the right of the plate	38	9,448
4	pick up the book and place it in the back compartment of the caddy	33	6,167
5	put both the alphabet soup and the tomato sauce in the basket	41	11,955
6	put the white mug on the left plate and put the yellow and white mug on the right plate	47	12,132
7	turn on the stove and put the moka pot on it	44	11,692
8	put the yellow and white mug in the microwave and close it	40	11,757
9	put the black bowl in the bottom drawer of the cabinet and close it	40	9,719

The source numeric Parquet is 17.70 MiB. The original videos/ tree contains 1,668 MP4 files (1.414 GiB), 417 per stream across four 256×256 RGB streams (AV1, 20 fps, no audio):

• observation.images.image — 417 MP4 files
• observation.images.wrist_image — 417 MP4 files
• observation.images.frontview_image — 417 MP4 files
• observation.images.sideview_image — 417 MP4 files

Converted Files

• TsFile: data/max_chr_libero_plus_10_add_object_all.tsfile
• Table: max_chr_libero_plus_10_add_object_all
• Rows: 112,885; episodes/devices: 417
• TsFile size: 12.51 MiB
• meta/ is mirrored; meta/info.json documents the converted schema, source revision, mapping, and video policy.

TsFile Schema

Time is INT64 milliseconds computed as round(timestamp * 1000); it restarts per episode. The source timestamp column is dropped as redundant with Time / 1000.

TAG columns (TsFile table-model device dimensions):

• episode_index
• task_index

Scalar FIELD columns:

• frame_index
• sample_index
• observation_state_0
• observation_state_1
• observation_state_2
• observation_state_3
• observation_state_4
• observation_state_5
• observation_state_6
• observation_state_7
• observation_states_ee_state_0
• observation_states_ee_state_1
• observation_states_ee_state_2
• observation_states_ee_state_3
• observation_states_ee_state_4
• observation_states_ee_state_5
• observation_states_joint_state_0
• observation_states_joint_state_1
• observation_states_joint_state_2
• observation_states_joint_state_3
• observation_states_joint_state_4
• observation_states_joint_state_5
• observation_states_joint_state_6
• observation_states_gripper_state_0
• observation_states_gripper_state_1
• action_0
• action_1
• action_2
• action_3
• action_4
• action_5
• action_6

Flattened FLOAT FIELD groups (full source prefix retained; . becomes _):

• observation.state (8 values) -> observation_state_0 ... observation_state_7
• observation.states.ee_state (6 values) -> observation_states_ee_state_0 ... observation_states_ee_state_5
• observation.states.joint_state (7 values) -> observation_states_joint_state_0 ... observation_states_joint_state_6
• observation.states.gripper_state (2 values) -> observation_states_gripper_state_0 ... observation_states_gripper_state_1
• action (7 values) -> action_0 ... action_6

Scalar renames and dropped fields:

• index -> sample_index
• Dropped after Time synthesis: timestamp

Encoding and Conversion Notes

• FLOAT/DOUBLE: GORILLA + LZ4.
• INT32/INT64 and Time: TS_2DIFF + LZ4.
• episode_index and task_index use the TsFile TAG/device mechanism; no synthetic aliases are created.
• Source index is retained as sample_index.
• All five fixed-width numeric vectors are flattened to scalar FLOAT fields.
• The train split is merged into one table-model TsFile; filter by TAG values to select an episode or task.

Videos and Frame Alignment

Videos are not copied here. They remain in the original videos/ tree. Rows align with source frames through episode_index, frame_index, and mirrored meta/episodes/ timestamps and file indexes.

Validation

A local JSON/Markdown report records source/staged/TsFile row counts, schema, TAG columns, readback status, and output size. Expected rows: 112,885.

Minimal Read Example

from tsfile import TsFileReader
reader = TsFileReader("data/max_chr_libero_plus_10_add_object_all.tsfile")
with reader.query_table(
    "max_chr_libero_plus_10_add_object_all",
    ["episode_index", "task_index", "frame_index", "sample_index", "action_0", "observation_state_0"],
    batch_size=65536,
) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()