HuskyMango/eval_act-5-22

该数据集由LeRobot创建，是一个用于机器人学习的多模态数据集。它基于单臂Franka机器人，包含5个episodes、2344帧数据，以parquet和视频文件格式存储。数据集特征包括动作控制数据（8维向量，表示位置、四元数旋转和夹爪状态）、观测状态（8维关节角度和夹爪状态），以及来自三个摄像头（cam_3_wrist、cam_4_wrist、cam_2_scene）的视频观测，每个视频分辨率为224x224，帧率为20fps。数据用于训练和评估机器人任务，如动作预测或控制。

This dataset was created using LeRobot and is a multimodal dataset for robotics learning. It is based on a single-arm Franka robot, containing 5 episodes and 2344 frames of data, stored in parquet and video file formats. The dataset features include action control data (an 8-dimensional vector representing position, quaternion rotation, and gripper state), observation state (8-dimensional joint angles and gripper state), and video observations from three cameras (cam_3_wrist, cam_4_wrist, cam_2_scene), each with a resolution of 224x224 and a frame rate of 20fps. The data is intended for training and evaluating robotics tasks, such as action prediction or control.