AiSaurabhPatil/quest3-acone

Quest3 Acone

This dataset contains teleoperated demonstrations for an Acone robot, recorded through a Quest 3 streaming setup and converted to the LeRobot dataset v3.0 format.

Dataset Details

• Format: LeRobot v3.0
• Robot type: acone
• Episodes: 200
• Frames: 111,458
• FPS: 30
• Task: sort nuts and bolts in different bins
• Video observations: 3 camera streams
• Action/state dimensions: 14 DoF

Observations and Actions

The dataset includes:

• observation.state: 14-dimensional robot state
• action: 14-dimensional action target
• observation.images.head_camera: 224x224 RGB video
• observation.images.left_wrist_camera: 224x224 RGB video
• observation.images.right_wrist_camera: 224x224 RGB video
• Standard LeRobot indexing fields: timestamp, frame_index, episode_index, index, task_index

The joint/action names are:

left_joint1, left_joint2, left_joint3, left_joint4, left_joint5, left_joint6, left_gripper, right_joint11, right_joint12, right_joint13, right_joint14, right_joint15, right_joint16, right_gripper.

Loading

from lerobot.datasets import LeRobotDataset

dataset = LeRobotDataset("AiSaurabhPatil/quest3-acone")
print(dataset.num_episodes, dataset.num_frames)

Provenance

The source dataset was stored locally in LeRobot v2.1 format and converted to v3.0 with:

python src/lerobot/scripts/convert_dataset_v21_to_v30.py \
  --repo-id=local/quest3-acone \
  --root=datasets/quest3-acone \
  --push-to-hub=false