key-unlock-v3.2

该数据集由LeRobot创建，专注于机器人技术领域，具体针对雅马哈双手机器人（yam_bimanual）。数据集包含200个episodes，总计212673帧数据，覆盖1个任务，并包含600个视频文件。数据以30fps的帧率采集，存储为parquet格式和mp4视频。数据集结构包括：observation.state（32维特征，涵盖左右机械臂的3D位置、6D旋转表示、夹爪状态和关节角度）、action（与observation.state维度相同的动作数据）、observation.images（来自三个相机视角的视频：头部、左腕和右腕，每个视频为3通道、360x640分辨率），以及时间戳、帧索引、episode索引等元数据。所有数据用于机器人学习任务，如模仿学习或强化学习，但具体应用场景未在README中明确说明。

This dataset was created by LeRobot and focuses on the field of robotics, specifically targeting the Yamaha bimanual robot (yam_bimanual). It contains 200 episodes, totaling 212673 frames of data, covering 1 task, and includes 600 video files. The data is collected at 30fps and stored in parquet format and mp4 videos. The dataset structure includes: observation.state (32-dimensional features, covering the 3D position, 6D rotation representation, gripper state, and joint angles of both robotic arms), action (action data with the same dimensions as observation.state), observation.images (videos from three camera perspectives: head, left wrist, and right wrist, each video with 3 channels and 360x640 resolution), as well as metadata such as timestamps, frame indices, and episode indices. All data is intended for robot learning tasks, such as imitation learning or reinforcement learning, but the specific application scenario is not explicitly stated in the README.