gsethia08/part1-deal_gather_cards

该数据集是一个用于机器人学习的数据集，专门针对手部机器人（robot_type: hand）设计。它包含1502个训练集（total_episodes: 1502），总计864,266帧（total_frames: 864266），覆盖19个不同任务（total_tasks: 19）。数据以30帧每秒（fps: 30）的速率采集，并以Parquet文件格式存储。数据集结构包括多个特征：观察数据（如相机图像和状态信息）、动作数据、语言指令、时间戳、帧索引、集索引等。相机图像分辨率为1080x1920（observation.camera.ego: shape [1080, 1920, 3]），状态数据包含48维浮点数（observation.state: shape [48]），代表左右手腕和手指的坐标及旋转信息。动作数据也是48维浮点数（action: shape [48]）。数据集使用Apache 2.0许可证（license: apache-2.0），主要用于机器人控制和模仿学习任务。This dataset is designed for robotics learning, specifically for hand-type robots (robot_type: hand). It contains 1502 training episodes (total_episodes: 1502) with a total of 864,266 frames (total_frames: 864266), covering 19 different tasks (total_tasks: 19). Data is captured at 30 frames per second (fps: 30) and stored in Parquet file format. The dataset structure includes multiple features: observations (such as camera images and state information), actions, language instructions, timestamps, frame indices, episode indices, etc. Camera images have a resolution of 1080x1920 (observation.camera.ego: shape [1080, 1920, 3]), state data consists of 48-dimensional floating-point numbers (observation.state: shape [48]) representing coordinates and rotations of left and right wrists and fingers. Action data is also 48-dimensional floating-point numbers (action: shape [48]). The dataset is licensed under Apache 2.0 (license: apache-2.0) and is primarily used for robot control and imitation learning tasks.