DAVIAN-Robotics/dh_tasks

该数据集是使用LeRobot创建的机器人学习数据集，专门用于机器人任务。数据集包含100个episodes，总计31283帧，涵盖8个不同的任务，使用Franka机器人平台。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20fps。数据集包括状态观测（如29维的关节位置、速度、末端执行器位姿、夹爪状态等）、动作（8维的末端执行器位置、姿态和夹爪控制）、以及来自三个摄像头的图像观测（左后、手腕、右后，每个摄像头提供3通道、480x640分辨率的RGB视频）。此外，还包含时间戳、帧索引、episode索引、任务索引等元数据。数据集仅提供训练分割（episodes 0到100），适用于机器人控制、模仿学习或强化学习任务。This dataset is a robotics learning dataset created using LeRobot, specifically designed for robot tasks. It contains 100 episodes, totaling 31,283 frames, covering 8 different tasks, and uses the Franka robot platform. The data is stored in parquet format, with a total data file size of 100MB and video file size of 200MB, at a frame rate of 20fps. The dataset includes state observations (such as 29-dimensional joint positions, velocities, end-effector poses, gripper state, etc.), actions (8-dimensional end-effector position, orientation, and gripper control), and image observations from three cameras (left back, wrist, right back, each providing 3-channel, 480x640 resolution RGB video). Additionally, it contains metadata such as timestamps, frame indices, episode indices, and task indices. The dataset only provides a training split (episodes 0 to 100) and is suitable for robot control, imitation learning, or reinforcement learning tasks.