jio2/fold_bag
收藏资源简介:
该数据集是一个机器人操作数据集,使用LeRobot框架创建。数据集包含71个训练片段,总计187,086帧数据,帧率为30fps。数据特征包括:12维动作向量(对应左右机械臂各6个关节的位置控制)、12维观测状态(机械臂关节位置)、以及三个摄像头视角的视频观测(左左腕、左上、右右腕视角,均为720x1280分辨率RGB视频)。此外,还包含时间戳、帧索引、片段索引、任务索引等元数据。数据集专为双臂跟随机器人(bi_so_follower)设计,适用于机器人学习任务,数据以Parquet格式和MP4视频格式存储,总数据量约300MB。
This dataset is a robotics manipulation dataset created using the LeRobot framework. It contains 71 training episodes with a total of 187,086 frames at 30fps. The features include: 12-dimensional action vectors (corresponding to position control for 6 joints each of the left and right robotic arms), 12-dimensional observation states (arm joint positions), and video observations from three camera perspectives (left left wrist, left top, right right wrist, all 720x1280 resolution RGB videos). Additionally, it includes metadata such as timestamps, frame indices, episode indices, and task indices. The dataset is designed for a bi_so_follower (bilateral follower) robot and is suitable for robot learning tasks. Data is stored in Parquet and MP4 video formats, with a total size of approximately 300MB.




