ecappiell/weighted_boxes_experiment_20260529_093939
收藏资源简介:
该数据集由LeRobot创建,专注于机器人学领域,用于模拟或记录机器人操作任务。数据集包含5个总片段,3550个总帧,1个总任务,数据以Parquet格式存储,视频以MP4格式存储。特征包括动作(6维浮点数组,控制机器人的肩部、肘部、腕部和夹爪位置)、观测状态(111维浮点数组,包含关节位置、电流、负载、速度、扭矩等详细信息)、两个图像观测(顶部和角落视角的480x640x3视频,帧率为30fps),以及时间戳、帧索引、片段索引、任务索引等元数据。数据集结构支持训练分割(0到5),适用于机器人控制、行为克隆或强化学习研究。
This dataset was created using LeRobot and focuses on robotics, designed for simulating or recording robot manipulation tasks. It includes 5 total episodes, 3550 total frames, and 1 total task, with data stored in Parquet format and videos in MP4 format. Features consist of actions (6-dimensional float32 array controlling robot shoulder, elbow, wrist, and gripper positions), observation states (111-dimensional float32 array with detailed joint positions, currents, loads, velocities, torques, etc.), two image observations (top and corner view videos of 480x640x3 resolution at 30fps), along with metadata such as timestamp, frame index, episode index, and task index. The dataset structure supports a train split (0 to 5) and is suitable for research in robot control, behavior cloning, or reinforcement learning.




