leokswang/20122025-foldclo-03_lerobot_format
收藏资源简介:
该数据集使用LeRobot工具创建,专注于机器人学领域,具体针对Yam机器人。数据集包含10个episodes,总计24888帧,帧率为30fps,所有数据均用于训练分割。数据以parquet格式存储,每个episode分为块,块大小为1000。数据集特征包括来自三个视角(顶部、左侧、右侧)的观测图像,每个图像分辨率为360x640x3;观测状态为14维浮点向量;动作为14维浮点向量;以及时间戳、帧索引、episode索引、索引和任务索引等元数据。数据集结构支持机器人控制和强化学习任务,但未提供主页、论文或引用信息。
This dataset was developed using the LeRobot toolkit, targeting the robotics domain with a specific focus on the Yam robot. It comprises 10 episodes, totaling 24,888 frames at 30fps, and all data is utilized for training segmentation tasks. The data is stored in Parquet format, with each episode divided into chunks of 1000. Dataset features include observed images from three viewpoints (top, left, right), each with a resolution of 360×640×3; observation states are 14-dimensional floating-point vectors; actions are 14-dimensional floating-point vectors; alongside metadata including timestamps, frame indices, episode indices, indices, and task indices. The dataset structure supports robotic control and reinforcement learning tasks, but no homepage, accompanying paper, or citation information is provided.




