fhliang/so101_pour_coffee
收藏资源简介:
该数据集由LeRobot创建,专注于机器人模仿学习任务,具体为“倒水”(pour)任务。数据集包含78个完整任务序列(episodes),总计49671帧数据,帧率为30fps。数据特征包括机器人的动作(如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置,共6个浮点数值)、观测状态(与动作相同的6个关节位置)、场景图像(480x640分辨率的RGB视频,编码为av1格式,无音频)、以及时间戳、帧索引、任务索引等元数据。数据集以parquet文件(用于结构化数据)和mp4视频文件存储,适用于机器人控制、行为克隆和模仿学习研究。
This dataset was created using LeRobot and focuses on robotics imitation learning, specifically the pour task. It consists of 78 episodes with a total of 49,671 frames at 30 fps. Features include robot actions (6 joint positions: shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), observation states (same 6 joint positions), scene images (480x640 RGB videos encoded in av1 format, without audio), and metadata such as timestamps, frame indices, and episode indices. The data is stored in parquet files for structured data and mp4 files for videos, suitable for robot control, behavior cloning, and imitation learning research.




