morealcholplz/vanilla-eval-50trial-seed1000

该数据集是一个用于机器人学习的数据集，由LeRobot创建。它包含500个任务集，总计149112帧数据，使用PandaOmron机器人类型。数据集以20帧每秒的速率采集，包括训练集（0:500）。数据特征涵盖观察图像（如机器人左代理视图、右代理视图和手眼视图），每个图像为256x256像素的RGB视频；观察状态向量（16维，包括基础位置、旋转、末端执行器相对位置和旋转，以及夹爪手指状态）；动作向量（12维，包括基础速度、躯干高度、控制模式、末端执行器位置和旋转变化，以及夹爪闭合状态）；以及奖励、完成标志、时间戳和索引信息。数据集结构以Parquet文件格式存储，适用于机器人强化学习或模仿学习任务。

This dataset is designed for robotics learning, created using LeRobot. It contains 500 task episodes with a total of 149,112 frames, utilizing the PandaOmron robot type. The data is captured at 20 frames per second and includes a training split (0:500). Features encompass observation images (e.g., robot left agent view, right agent view, and eye-in-hand view), each as RGB video frames with 256x256 resolution; observation state vectors (16-dimensional, including base position, rotation, end-effector relative position and rotation, and gripper finger states); action vectors (12-dimensional, including base velocities, torso height, control mode, end-effector position and rotation changes, and gripper closure); along with reward, done flags, timestamps, and index information. The dataset structure is stored in Parquet format, suitable for robotics reinforcement learning or imitation learning tasks.