rishgang22/cube_stacking

该数据集是一个机器人任务数据集，专注于方块堆叠任务，使用LeRobot框架创建。数据集包含101个总episodes和130586个总frames，数据以30fps的帧率采集。特征包括动作（action）和观察状态（observation.state），其中动作由6个浮点数值组成，对应机器人关节位置（如肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）。观察状态同样包含6个浮点数值，表示机器人状态。此外，数据集提供来自overhead（头顶）和wrist（手腕）的图像视频观察，每个视频分辨率为480x640，3个通道，使用av1编解码器。其他元数据包括时间戳、帧索引、episode索引、索引和任务索引。机器人类型为so_follower，数据以parquet文件（总大小100MB）和视频文件（总大小200MB）形式存储，分割为训练集（涵盖所有episodes）。This dataset is a robotics task dataset focused on cube stacking, created using the LeRobot framework. It contains a total of 101 episodes and 130,586 frames, collected at a frame rate of 30 fps. Features include action and observation.state, where action consists of 6 float32 values corresponding to robot joint positions (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions). Observation.state also includes 6 float32 values representing robot state. Additionally, the dataset provides image video observations from overhead and wrist cameras, each with a resolution of 480x640, 3 channels, using the av1 codec. Other metadata includes timestamp, frame index, episode index, index, and task index. The robot type is so_follower, and data is stored in parquet files (total size 100 MB) and video files (total size 200 MB), with splits for training (covering all episodes).