koch_grasp_tape_2

该数据集是使用LeRobot创建的，主要用于机器人相关任务。数据集包含50个集，总计29385帧，100个视频，1个任务。数据集的结构包括动作、观察状态、观察图像（自拍和jetbot）、时间戳、帧索引、集索引、索引和任务索引等特征。动作和观察状态的特征包括6个浮点型数据，分别代表主肩盘、主肩提升、主肘弯曲、主腕弯曲、主腕滚动和主夹持器。观察图像的特征包括480x640x3的视频数据，帧率为30fps。数据集的分割信息显示所有数据都用于训练。

This dataset was created using LeRobot, and is primarily intended for robotics-related tasks. It includes 50 episodes, totaling 29385 frames, 100 videos, and one single task. The dataset's structure encompasses features including actions, observation states, observation images (selfie and jetbot), timestamps, frame indices, episode indices, global index, and task index. Both actions and observation states feature 6 floating-point values, which correspond to main shoulder pan, main shoulder lift, main elbow flexion, main wrist flexion, main wrist roll, and main gripper, respectively. The observation images consist of 480x640x3 video data with a frame rate of 30 fps. Dataset split information indicates that all data is used for training.