act-training-v2-1

该数据集由LeRobot创建，主要用于机器人领域。数据集包含10个episodes，共5278帧，涉及1个任务。数据以parquet格式存储，总数据文件大小为100MB，视频文件大小为200MB，帧率为30fps。数据集包含多种特征，如动作（6个浮点数表示的关节位置）、观察状态（6个浮点数表示的关节位置）、手腕和前置摄像头图像（480x640x3的视频数据）、时间戳、帧索引、episode索引等。数据集结构在meta/info.json中有详细描述。

This dataset was developed by LeRobot, and is primarily targeted for robotics research. It contains 10 episodes totaling 5278 frames, covering a single task. The data is stored in Parquet format, with the total size of data files being 100 MB, while the video files have a total size of 200 MB. The frame rate of the dataset is 30 fps. The dataset includes various features, such as actions (6 floating-point numbers representing joint positions), observation states (6 floating-point numbers representing joint positions), wrist and front-facing camera images (480×640×3 video data), timestamps, frame indices, episode indices, and others. The detailed structure of the dataset is described in meta/info.json.