SharpaIT/Robotic_Origami_Challenge

Robotic Origami Challenge: Fold Plane Demonstrations

Real-world LeRobot demonstrations for dexterous paper-airplane folding.

Overview

Robotic Origami Challenge: Fold Plane Demonstrations is a real-world teleoperation dataset for folding a paper airplane with a bimanual dexterous robot system. It is released by Sharpa in a LeRobot-compatible format for the Robotic Origami Challenge community.

Origami is a demanding benchmark for embodied AI: paper is thin, deformable, easy to occlude, and highly sensitive to contact timing and crease quality. A successful policy must coordinate two arms, dexterous hands, tactile feedback, and multi-view vision over a long sequential horizon.

The dataset is designed for imitation learning, visuomotor policy learning, visual-tactile representation learning, long-horizon action modeling, and policy development for the Robotic Origami Challenge.

Task Traditional paper-airplane folding	Format LeRobot v3.0 / v2.1	Scale 51 seasons / 682 episodes	Frequency 30 FPS
Video 6 synchronized streams	State / Action 65D joint space	Tactile 10-fingertip 6-axis signals + tactile video	Use Training and policy development

Target Figure

The target figure for the challenge is a traditional paper airplane, or kami hikoki. Every team folds the same figure.

Target Figure Item	Specification
Figure	Traditional Japanese paper airplane
Paper size	15 x 15 cm
Paper weight	>= 60 gsm
Fold count	6 folds
Attempt limit	10 minutes
Success criterion	The judge panel declares whether the robot folded a recognizable airplane
Ranking criterion	Successful attempts are ranked by folding time
Judge	Origami Grand Master from the Nippon Origami Association

The task is judged on crease accuracy, structural fidelity, symmetry, and paper integrity. Whether the resulting plane flies is not part of the official score, but the folded structure should remain recognizable as the target airplane.

Challenge Context

The Robotic Origami Challenge is an IROS 2026 competition in Pittsburgh focused on dexterous robotic origami. It benchmarks fine-motor, sequential paper manipulation against an origami standard curated by expert artists.

Resource	Description
Teleoperation data	500+ demonstrations collected on the target folding sequence, together with fold-sequence diagrams and reference folds.
Simulation	NVIDIA Isaac Sim based environment with thin-shell paper physics, plastic creasing, fold memory, partner robotic-hand digital twins, and the same scoring rubric as the live event.
Real-world evaluation	A remote lab where participants upload policies and run them on the same bimanual arms and Sharpa Hands rig used for the challenge.

The remote evaluation setup lowers the hardware barrier for participants: teams can train in simulation, use the demonstration data, upload policies, and evaluate on a standardized real-world system.

Example Views

The demonstrations include synchronized head, wrist, and tactile video streams. Each preview below uses a 3-minute window from the same lerobotv2.1 episode, played at 10x speed. GIF previews render directly in Markdown; click any preview to open the MP4 version. Tactile previews preserve their full wide-frame layout.

Head Left	Head Right
Wrist Left	Wrist Right
Tactile Deformation	Raw Tactile

Dataset Capabilities

Capability	Dataset Support
Long-horizon imitation learning	Real-world demonstrations for the official paper-airplane folding sequence
Multi-view visuomotor policies	Synchronized head-camera and wrist-camera observations
Visual-tactile learning	High-resolution tactile videos, synchronized raw tactile camera views, and 10-fingertip 6-axis tactile signals
Joint-space control	65D synchronized state and action for two arms, two dexterous hands, and torso/motor-related joints
LeRobot ecosystem	Full lerobot3.0 coverage across all seasons, with lerobotv2.1 available for many seasons

Dataset Statistics

Item	Value
Total collection seasons	51
lerobot3.0 seasons	51
lerobot3.0 episodes	682
lerobot3.0 frames	4,763,267
lerobotv2.1 seasons	41
lerobotv2.1 episodes	555
lerobotv2.1 frames	3,893,595
FPS	30
Video streams	6
State/action dimension	65

For new users, we recommend starting from lerobot3.0, since it covers all 51 seasons in this release.

Get Started

Download The Dataset

Make sure Git LFS is installed before cloning from Hugging Face.

git lfs install
git clone https://huggingface.co/datasets/SharpaIT/Robotic_Origami_Challenge

If you want to clone only metadata first and fetch large files later:

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/datasets/SharpaIT/Robotic_Origami_Challenge

If you only want a specific season, use sparse checkout:

git init Robotic_Origami_Challenge
cd Robotic_Origami_Challenge
git remote add origin https://huggingface.co/datasets/SharpaIT/Robotic_Origami_Challenge
git sparse-checkout init
git sparse-checkout set season_POC22032_2026_05_14_19_21_01_train README.md
git pull origin main

Quick Inspection

Each season contains one or two LeRobot exports. Inspect meta/info.json first to understand the exact schema and file templates.

import json
from pathlib import Path

dataset_root = Path("Robotic_Origami_Challenge")
episode_root = dataset_root / "season_POC22032_2026_05_14_19_21_01_train" / "lerobot3.0"

with open(episode_root / "meta" / "info.json", "r") as f:
    info = json.load(f)

print(info["total_episodes"])
print(info["total_frames"])
print(info["features"].keys())

Dataset Structure

The dataset is organized by collection season. Each season may contain a lerobot3.0 export and, when available, a lerobotv2.1 export.

Robotic_Origami_Challenge/
├── README.md
├── season_POC22032_2026_05_14_19_21_01_train/
│   ├── lerobot3.0/
│   │   ├── meta/
│   │   │   ├── info.json
│   │   │   ├── modality.json
│   │   │   ├── episodes.jsonl
│   │   │   └── tasks.jsonl
│   │   ├── data/
│   │   │   └── chunk-000/
│   │   └── videos/
│   │       ├── observation.images.head_left/
│   │       ├── observation.images.head_right/
│   │       ├── observation.images.wrist_left/
│   │       ├── observation.images.wrist_right/
│   │       ├── observation.images.tactile_deform/
│   │       └── observation.images.tactile_raw/
│   └── lerobotv2.1/
│       ├── meta/
│       ├── data/
│       └── videos/
└── season_.../

LeRobot Storage Layout

Part	Description
meta/	Dataset metadata, feature schema, task metadata, and path templates
data/	Episode frame data stored as Apache Parquet files
videos/	Per-camera MP4 videos

The most important metadata file is meta/info.json. It defines total_episodes, total_frames, fps, splits, data_path, video_path, and features.

File Path Templates

LeRobot v3.0 uses templates similar to:

data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

LeRobot v2.1 uses templates similar to:

data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4

Features Schema

Main Feature Groups

Feature	Type	Shape	Description
observation.state	float32	65	Joint-space robot state
action	float32	65	Joint-space action
observation.state.joint_torque	float32	65	Joint torque signal
observation.tactile	float32	60	10-fingertip 6-axis force/torque tactile signal
observation.images.*	video	varies	Multi-view visual observations
timestamp	float32	1	Timestamp
frame_index	int64	1	Frame index within an episode
episode_index	int64	1	Episode index
task_index	int64	1	Task index

Proprioceptive State

The 65D observation.state and action vectors are ordered as follows:

Range	Names	Meaning
0 - 6	left_arm_j0 to left_arm_j6	Left arm joints
7 - 28	left_hand_j0 to left_hand_j21	Left dexterous hand joints
29 - 35	right_arm_j0 to right_arm_j6	Right arm joints
36 - 57	right_hand_j0 to right_hand_j21	Right dexterous hand joints
58 - 64	motor_j0 to motor_j6	Torso / motor-related joints

Video Streams

The complete visual observation set contains six camera streams:

Feature key	Description	Shape
observation.images.head_left	Left head camera	480 x 480 x 3
observation.images.head_right	Right head camera	480 x 480 x 3
observation.images.wrist_left	Left wrist camera	480 x 480 x 3
observation.images.wrist_right	Right wrist camera	480 x 480 x 3
observation.images.tactile_deform	Tactile deformation video	480 x 1200 x 3
observation.images.tactile_raw	Raw tactile video	480 x 1600 x 3

Video files are MP4 without audio. Codec may differ across exports and seasons. lerobot3.0 is primarily AV1, while lerobotv2.1 is primarily H.264.

Tactile Modality

The dataset includes tactile observations as both compact numeric signals and high-resolution video streams. These modalities are synchronized with the robot state, action, and visual camera streams at 30 FPS, making them suitable for contact-rich policy learning and visual-tactile representation learning.

Tactile feature	Type	Shape	Description
observation.tactile	float32	60	Per-frame tactile force/torque signal: 10 fingertips x 6 axes
observation.images.tactile_deform	video	480 x 1200 x 3	Deformation-oriented tactile video stream that visualizes contact-induced surface changes
observation.images.tactile_raw	video	480 x 1600 x 3	Raw tactile camera stream preserving the full tactile sensor image layout

The observation.tactile vector provides a compact force/torque representation for each frame. It contains 10 fingertip groups: left and right thumb, index, middle, ring, and little. Each fingertip contributes six values, ordered as fx, fy, fz, tx, ty, and tz, for a total of 60 dimensions. This makes it suitable for models that consume structured numeric tactile feedback alongside proprioception and action labels. The two tactile video streams provide complementary image-based tactile observations: tactile_deform emphasizes deformation patterns caused by contact, while tactile_raw preserves the raw tactile image for users who want to build their own visual-tactile preprocessing or representation learning pipeline.

For downstream experiments, users can start with observation.tactile as a lightweight contact signal, then add one or both tactile video streams when the model architecture can handle the additional spatial resolution and bandwidth. The per-season meta/info.json records the exact feature schema and channel names.

Usage Recommendations

This release is provided as training data. The official Robotic Origami Challenge evaluation is performed through the competition's real-world evaluation setup rather than through a fixed validation split in this repository.

For local experiments, users may split by season to avoid mixing demonstrations from the same collection session across train and evaluation sets.

For policy learning, a typical setup is:

• Visual observations: one or more observation.images.* streams
• Proprioception: observation.state
• Optional tactile inputs: observation.tactile, observation.images.tactile_deform, and/or observation.images.tactile_raw
• Supervision target: action

For multi-view policies, start with:

observation.images.head_left
observation.images.head_right
observation.images.wrist_left
observation.images.wrist_right

Then add tactile video streams if your model can use high-resolution tactile observations:

observation.images.tactile_deform
observation.images.tactile_raw

Dataset Notes

• The dataset is released by season, not as a single flattened LeRobot root.
• lerobot3.0 has full season coverage in this release.
• Some seasons also include lerobotv2.1 for compatibility with older pipelines.
• motor_j0 to motor_j6 are the torso/motor-related dimensions in observation.state and action.
• The task is contact-rich and deformable: policies should expect paper motion, occlusions, hand-paper contacts, and fine-grained crease formation.

License and Terms

This dataset is released under the Creative Commons Attribution 4.0 International License (CC-BY-4.0). You may use, share, and adapt the dataset, including for commercial purposes, provided that you give appropriate attribution.

If you use the dataset for Robotic Origami Challenge participation, please also follow the official competition rules and evaluation protocol.

Citation

If this dataset contributes to your research, please cite or acknowledge the dataset and the Robotic Origami Challenge.

@misc{robotic_origami_challenge_fold_plane_2026,
  title        = {Robotic Origami Challenge Fold Plane LeRobot Dataset},
  howpublished = {\url{https://huggingface.co/datasets/SharpaIT/Robotic_Origami_Challenge}},
  year         = {2026}
}

@misc{robotic_origami_challenge_2026,
  title        = {The Robotic Origami Challenge},
  howpublished = {\url{https://robotic-origami-challenge.github.io/}},
  year         = {2026}
}