pig3_seg (TsFile)

Apache TsFile version of adrienloizeau/pig3_seg, a LeRobot v2.1 demonstration dataset recorded with an SO-100 arm.

Overview

This dataset was created using LeRobot. It records one SO-100 manipulation task with per-frame joint state and action vectors, along with three synchronized camera streams (robot, webcam, seg).

• Robot: SO-100 (so100)
• Scale: 50 episodes, 26,799 frames, 1 task, 30 fps
• Split: train

Schema (TsFile structure)

The TsFile table is named pig3_seg.

Role	Columns
Time	Time, INT64 milliseconds
TAG	episode_index, task_index
FIELD	frame_index, sample_index
FIELD	action_0 … action_5, FLOAT
FIELD	observation_state_0 … observation_state_5, FLOAT

The original 6-element action and observation.state vectors are flattened into scalar FLOAT measurements. The source index column is retained as sample_index.

Conversion notes

• Time = round(timestamp * 1000) with millisecond precision; time restarts inside each episode, while episode_index and task_index identify the TsFile device.
• The original timestamp field is omitted because it is exactly represented by Time / 1000.
• The three camera streams (observation.images.robot, observation.images.webcam, observation.images.seg) are video and are NOT included in this repository; they remain in the source videos tree.
• meta/ is mirrored from the source. Aside from the redundant timestamp column and the excluded videos, no source rows or numeric fields are dropped.

Read example

from tsfile import TsFileReader

with TsFileReader("data/pig3_seg.tsfile") as reader:
    print(reader.get_all_table_schemas()["pig3_seg"])

Source & license

• Original dataset: https://huggingface.co/datasets/adrienloizeau/pig3_seg
• Author: adrienloizeau
• License: apache-2.0