fecasado/burger-to-plate
收藏资源简介:
该数据集是一个机器人任务数据集,使用LeRobot工具创建,专门用于机器人学习和控制任务。数据集包含93个episodes,总计137,976帧,采样频率为15 FPS。数据来源于blueberry_ros机器人类型,涉及单一任务。数据集结构包括动作特征(26维浮点向量,涵盖左右手臂的线性和角速度、手指关节控制等)、观测状态特征(55维浮点向量,包括左右手臂和手指关节的位置和力矩,以及视线坐标)、多视角图像观测(左、右、用户和用户视线视角的视频,分辨率480x640,3通道RGB,AV1编码),以及时间戳、帧索引等元数据。所有数据以Parquet格式存储,总数据文件大小为100 MB,视频文件大小为200 MB。数据集仅包含训练集(splits: train),适用于机器人模仿学习或强化学习研究。
This dataset is a robotics task dataset created using the LeRobot tool, specifically designed for robot learning and control tasks. It contains 93 episodes with a total of 137,976 frames, sampled at 15 FPS. The data originates from the blueberry_ros robot type and involves a single task. The dataset structure includes action features (a 26-dimensional float32 vector covering linear and angular velocities of left and right arms, finger joint controls, etc.), observation state features (a 55-dimensional float32 vector including positions and efforts of left and right arm and finger joints, as well as gaze coordinates), multi-view image observations (videos from left, right, user, and user gaze perspectives with 480x640 resolution, 3 RGB channels, AV1 encoding), and metadata such as timestamp and frame index. All data is stored in Parquet format, with a total data file size of 100 MB and video file size of 200 MB. The dataset includes only a training set (splits: train) and is suitable for research in robot imitation learning or reinforcement learning.




