kunhsiang/eval_exp9_sc2var_grasp_the_small_box_20260527-170521
收藏资源简介:
该数据集是通过LeRobot平台创建的机器人学数据集,专注于一个具体的任务:抓取小盒子。数据集包含1个完整的episode,总计2384帧数据,帧率为30fps。数据以parquet文件格式存储,并包含对应的mp4格式视频文件。数据集的特征包括:动作(action),为6维浮点数组,对应机器人肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪的位置;状态观测(observation.state),同样为6维浮点数组,反映机器人各关节的实时位置;图像观测(observation.images),提供来自三个摄像头视角的视频数据,包括前视角(front)、顶视角(top)和夹爪视角(gripper),每个视频的分辨率为480x640,3通道彩色,编码格式为av1。此外,数据集还包含时间戳(timestamp)、帧索引(frame_index)、episode索引(episode_index)、数据索引(index)和任务索引(task_index)等元数据。数据集适用于机器人控制、模仿学习或强化学习的研究与应用。
This dataset is a robotics dataset created using the LeRobot platform, focusing on a specific task: grasping a small box. It contains 1 complete episode with a total of 2384 frames at a frame rate of 30fps. The data is stored in parquet format and includes corresponding video files in mp4 format. The dataset features include: action, a 6-dimensional float32 array corresponding to the positions of shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper; state observation (observation.state), also a 6-dimensional float32 array reflecting the real-time joint positions of the robot; image observations (observation.images), providing video data from three camera perspectives: front, top, and gripper, each with a resolution of 480x640, 3 color channels, and encoded in av1 format. Additionally, the dataset includes metadata such as timestamp, frame_index, episode_index, index, and task_index. This dataset is suitable for research and applications in robot control, imitation learning, or reinforcement learning.




