Datasets:

jzpeng
/

ajaysri_long_cable_insertion_2_raw_plus_dagger2x

license: apache-2.0
authors:
  - Ajay Sridhar
task_categories:
  - robotics
tags:
  - tsfile
  - time-series
  - timeseries
  - robotics
  - lerobot
  - arx
  - teleop
  - long-cable
  - subtask
  - dagger
  - v3
modality:
  - timeseries
  - tabular
pretty_name: Long Cable Insertion 2 Raw Plus DAgger 2x (TsFile)
configs:
  - config_name: default
    data_files:
      - split: train
        path: data/long_cable_insertion_2_raw_plus_dagger2x_train.tsfile
size_categories:
  - 10K<n<100K

Long Cable Insertion 2 Raw Plus DAgger 2x (TsFile)

This dataset is an Apache TsFile conversion of ajaysri/long_cable_insertion_2_raw_plus_dagger2x, a LeRobot v3 ARX robot-manipulation dataset for long-cable routing subtasks.

Modalities: Time-series. The converted repository contains numeric robot state, velocity, action, timing, episode/task tags, and mirrored source metadata. Camera videos remain in the original Hugging Face dataset.

Source Dataset and Author

Original dataset: ajaysri/long_cable_insertion_2_raw_plus_dagger2x
Pinned source revision: e539e4d3a54766126424675ec3abe08f2414a60b
Original author, repository owner, and uploader: Ajay Sridhar (ajaysri)
License: Apache-2.0
Robot type: arx
LeRobot codebase version: v3.0
Sampling rate: 50 fps
Split: train, episodes 0 through 537
Source scale: 538 episodes, 70,596 frames, 3 tasks
Source frame layout: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
Source frame shards: 538
Source episode-metadata shards: 538
Paper, homepage, and completed citation: not provided by the source card

The source provenance manifest reports 338 routed subtask episodes extracted from 114 candidate raw HDF5 episodes, plus 200 DAgger episodes made as two copies of 100 source DAgger episodes. One raw episode was rejected during source-dataset creation before publication because its switch sequence failed segmentation validation; this TsFile conversion itself drops no published rows.

Tasks and Scale

`task_index`	Task	Episodes	Rows
`0`	route the cable through the center router	179	22,778
`1`	route the cable through the right router	185	24,715
`2`	route the cable through the left router	174	23,103

The converted train split contains 70,596 rows across 538 episode devices.

Converted Files

TsFile: data/long_cable_insertion_2_raw_plus_dagger2x_train.tsfile
Table: long_cable_insertion_2_raw_plus_dagger2x_train
Rows: 70,596
TsFile size: 10.93 MiB
Time precision: milliseconds
Metadata: meta/ is mirrored from the source, with meta/info.json rewritten to document the TsFile schema, source revision, conversion mapping, and video policy.

Compact Encoding Profile

The TsFile is written with the native table writer and the following explicit type-aware encodings to keep the binary smaller than the source frame Parquet:

FLOAT, DOUBLE: GORILLA + LZ4
INT32, INT64: TS_2DIFF + LZ4
Time: TS_2DIFF + LZ4
BOOLEAN: RLE + LZ4 (no BOOLEAN FIELD occurs in this dataset)
TAG: TsFile table TAG/device segments; tag values are not duplicated as ordinary FIELD measurements

TsFile Schema

Time is an INT64 millisecond timestamp computed as round(timestamp * 1000). It restarts within each episode.

TAG columns:

episode_index — source episode number, 0 through 537
task_index — source task number, 0 through 2

Scalar FIELD columns:

frame_index — frame number within an episode
sample_index — the source global index column, renamed

Flattened FLOAT FIELD groups:

observation.state[14] -> observation_state_0 ... observation_state_13
observation.velocity[14] -> observation_velocity_0 ... observation_velocity_13
action[14] -> action_0 ... action_13

The table has 47 columns including Time: 2 TAG columns, 44 FIELD columns, and the TIME column.

Conversion Notes

The shared config-driven lerobot converter is used with its native Python table writer; the dataset-specific Python file is a thin local orchestration and documentation entry point.
All 538 source frame Parquet files in the train split are merged into one table-model TsFile. Filter by episode_index and task_index to select a trajectory or routing subtask.
Vector columns are fully flattened into scalar fields. Full source prefixes are preserved and . is replaced with _.
The source timestamp column is the only published numeric source column not stored separately. It is redundant after the exact millisecond Time mapping (timestamp = Time / 1000 seconds at the source's 50 Hz cadence).
The source index column is retained as sample_index; frame_index is retained unchanged.
No published frames, episodes, tasks, state dimensions, velocity dimensions, or action dimensions are intentionally dropped.

Videos

Videos are not duplicated in the converted repository. The pinned source has 1,614 frame-aligned MP4 files in three streams:

videos/observation.images.video_left/ — 538 MP4 files
videos/observation.images.video_overhead/ — 538 MP4 files
videos/observation.images.video_right/ — 538 MP4 files

The source template is videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4. Each stream contains one file for each of the 538 episodes. Numeric TsFile rows remain aligned with the original videos through episode_index, frame_index, and the source metadata under meta/episodes/.

Validation

The generated TsFile is checked for successful native-writer completion, non-zero size, table schema, and row-count equality with the staged Parquet. The expected and validated row count is 70,596; the source contains no duplicate (episode_index, task_index, Time) keys after millisecond conversion.

Minimal Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/long_cable_insertion_2_raw_plus_dagger2x_train.tsfile")
print(reader.get_all_table_schemas()["long_cable_insertion_2_raw_plus_dagger2x_train"])
reader.close()

Source and License

The source dataset is published by Ajay Sridhar (ajaysri) under the Apache License 2.0 and was created with LeRobot. The source card does not provide a paper or completed citation, so none is inferred here.