so_101_pick_and_place_candy_20260618_174837

该数据集由LeRobot创建，专注于机器人学任务，具体涉及pick and place操作（基于数据集名称推断）。数据集包含10个episodes，总计5089帧数据，以30fps的帧率记录。数据特征包括机器人的动作（如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部滚动和夹爪位置）、观察状态（与动作相同的关节位置）、固定相机和手眼相机的视频图像（分辨率为480x640，3通道，编码为av1），以及时间戳、帧索引、episode索引等元数据。机器人类型为so_follower，数据集用于训练目的（分割为train集），数据以parquet文件格式存储，视频以mp4格式存储，总数据大小约为100MB，视频大小约为200MB。数据集结构支持机器人控制和视觉感知研究。

This dataset was developed by LeRobot, focused on robotics tasks, with a particular focus on pick and place operations (inferred from the dataset's name). The dataset contains 10 episodes, totaling 5089 frames captured at a frame rate of 30 fps. The data features include robot control actions (such as shoulder translation, shoulder lift, elbow flexion, wrist flexion, wrist rotation, and gripper position), observation states (consisting of identical joint positions to the recorded actions), video images from fixed cameras and eye-in-hand cameras (with a resolution of 480×640, 3 color channels, encoded in AV1), along with metadata including timestamps, frame indices, and episode indices. The target robot platform is so_follower. This dataset is intended for model training and has been split into a training set, with the tabular motion and metadata data stored in Parquet format, while the video footage is saved in MP4 format. The total size of the Parquet-stored data is approximately 100 MB, and the video files occupy roughly 200 MB in total. The structure of this dataset facilitates research in robot control and visual perception.