TRI-ML/robosuite_mg
收藏资源简介:
该数据集是一个机器人技术相关的多任务学习数据集,使用LeRobot创建。数据集包含来自agentview和wrist摄像头的视频数据(分辨率为84x84,3通道,帧率20fps),以及32维的状态向量和7维的动作向量。数据集总共有5400个episodes,810000帧,分为2个任务。数据以parquet格式存储,视频以mp4格式存储。数据集结构详细描述了每个特征的形状和类型,适用于机器人控制和多任务学习的研究。
This is a multi-task learning dataset related to robotics, created using LeRobot. The dataset contains video data from agentview and wrist cameras with a resolution of 84×84, 3 channels, and a frame rate of 20fps, as well as 32-dimensional state vectors and 7-dimensional action vectors. In total, the dataset consists of 5400 episodes and 810000 frames, which are divided into two tasks. The data is stored in Parquet format, while the videos are stored in MP4 format. The dataset structure elaborates on the shape and type of each feature, and is suitable for research on robotic control and multi-task learning.




