DrawTriangle v1 TsFile
Apache TsFile representation of brandonyang/DrawTriangle-v1, a LeRobot v2.1 robotics dataset.
Source
- Author and publisher: Brandon Yang (
brandonyang) - License: Apache-2.0
- Task: Draw a triangle on the canvas using the robot's end-effector.
- Split:
train(0:1000) - Sampling rate: 20 fps
- Episodes: 1,000; frame rows: 230,792; tasks: 1; source Parquet shards: 1,000
- Source frame layout:
data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
The source card does not provide a paper or completed BibTeX citation.
TsFile layout
The numeric rows are stored in data/brandonyang_drawtriangle_v1.tsfile as table brandonyang_drawtriangle_v1. Source metadata is retained under meta/; source Parquet files are not placed there.
Time = round(timestamp * 1000) is an INT64 millisecond timeline that restarts at zero for each episode. The source timestamp is dropped because it equals Time / 1000 seconds. index is renamed to sample_index, and frame_index is preserved.
| Role | Columns |
|---|---|
| TIME | Time (INT64, milliseconds) |
| TAG/device | episode_index, task_index |
| FIELD | frame_index, sample_index (INT64) |
| FIELD | action[6] -> action_0 ... action_5 (FLOAT) |
| FIELD | observation.state[7] -> observation_state_0 ... observation_state_6 (FLOAT) |
No numeric rows, action dimensions, or state dimensions are dropped.
Videos
The original repository contains 1,000 AV1 MP4 files for observation.images.base_camera under videos/chunk-000/observation.images.base_camera/:
videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4
Videos are not included in this TsFile dataset. Numeric rows remain aligned with the original videos through episode_index and frame_index.
Read example
from tsfile import TsFileReader
reader = TsFileReader("data/brandonyang_drawtriangle_v1.tsfile")
columns = ["episode_index", "task_index", "frame_index", "sample_index", "action_0", "observation_state_0"]
with reader.query_table("brandonyang_drawtriangle_v1", columns, batch_size=4096) as result:
print(result.read_arrow_batch().to_pandas().head())
reader.close()
- Downloads last month
- 49