eval_xvla_candy_sorting_ood
收藏资源简介:
该数据集是使用LeRobot平台创建的机器人学数据集,采用Apache 2.0许可证。它包含一个双臂机器人(类型为bi_yam_follower)执行任务时记录的多模态数据,规模为23个episodes,总计16014帧,覆盖4个不同的任务,所有数据均用于训练。数据以分块形式存储,每块包含1000个数据点,数据文件总大小约为100 MB,关联的视频文件总大小约为200 MB,视频帧率为30 fps。数据集提供了丰富的特征,包括机器人的动作指令(14维浮点数)、机器人状态观测(14维浮点数)、来自三个固定视角(右、左、顶部)的图像观测(均为360x640 RGB视频,使用AV1编解码器),以及时间戳、帧索引、episode索引、全局索引和任务索引等元数据。该数据集适用于机器人模仿学习、行为克隆、策略学习等研究任务。
This dataset is a robotics dataset created using the LeRobot platform under the Apache 2.0 license. It contains multimodal data recorded by a dual-arm robot (type bi_yam_follower) while performing tasks, with a scale of 23 episodes totaling 16014 frames, covering 4 different tasks, all data is used for training. The data is stored in chunks, each containing 1000 data points, with the total data file size approximately 100 MB and associated video files totaling about 200 MB, with a video frame rate of 30 fps. The dataset provides rich features, including robot action commands (14-dimensional floating-point numbers), robot state observations (14-dimensional floating-point numbers), image observations from three fixed perspectives (right, left, top, all in 360x640 RGB video using the AV1 codec), and metadata such as timestamps, frame indices, episode indices, global indices, and task indices. This dataset is suitable for research tasks such as robot imitation learning, behavior cloning, and policy learning.




