Datasets:

DaivdYuan
/

umi-bimanual-cloth-folding-lerobot

Parquet error: Scan size limit exceeded: attempted to read 911167144 bytes, limit is 300000000 bytes Make sure that 1. the Parquet files contain a page index to enable random access without loading entire row groups2. otherwise use smaller row-group sizes when serializing the Parquet files

Error code:   TooBigContentError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Bimanual Cloth Folding

Visualizers

Overview

This dataset is a conversion from an upstream robotics dataset into LeRobot v3-compatible format. It is intended to provide reproducible access in a unified schema.

Source Dataset

Dataset ID: umi-bimanual-cloth-folding
Project: UMI (Core)
Task: Bimanual Cloth Folding
Upstream download/source URL: https://real.stanford.edu/umi/data/bimanual_cloth_folding/bimanual_cloth_folding.zarr.zip
Upstream project page: https://umi-data.github.io
Contact: chuerpan@stanford.edu
Original paper: Chi et al. 2024
Notes: 249 demos; 1 env; image+proprio; bimanual

Dataset Format

Storage format: LeRobot v3 parquet-based dataset
Total episodes: 249
Total frames: 309476
FPS: 30
Robot type: generic_robot
Observation feature: observation.state with shape [14]
Action feature: action with shape [14]
Video streams: present

Video Modalities

observation.images.camera0_rgb: shape [224, 224, 3]
observation.images.camera1_rgb: shape [224, 224, 3]
Frame rate: 30 FPS

Conversion Notes

Target schema: LeRobot v3
Conversion path: ingest -> preprocess -> zarr probe -> convert 1 episode -> upload
action fallback may be used when upstream data/action is missing.

Dataset Structure

meta/info.json:

{
  "codebase_version": "v3.0",
  "robot_type": "generic_robot",
  "total_episodes": 249,
  "total_frames": 309476,
  "fps": 30,
  "splits": {"train": "0:249"},
  "data_path": "data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet",
  "video_path": "videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4",
  "features": {
    "observation.state": {"dtype": "float32", "shape": [14]},
    "action": {"dtype": "float32", "shape": [14]}
  }
}

Data Splits

{'train': '0:249'}

Citation

This dataset was created using LeRobot. Please also cite the original upstream project and paper listed above.

Attribution

Original dataset credit remains with upstream creators.
Conversion and curation contribution:
- Convertor: David Yuan, davidy02@stanford.edu, Stanford University
- NMotion Initiative (opensource community): we make embodied data accessible

Downloads last month: 101

Number of rows:

Total file size:

969 MB