may7_TRIMMED_first_50_frames_merged
收藏资源简介:
该数据集是一个机器人控制数据集,使用LeRobot框架创建。数据集包含271个训练episodes,总计231,525个数据帧,覆盖单一任务。数据以结构化格式存储,包含机械臂的动作指令、状态观测和视觉观测。具体而言,动作和状态观测均为6维浮点向量,分别对应机械臂的六个关节位置(肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)。视觉观测为来自前视摄像头的视频流,分辨率为1080x1920,帧率为30fps,采用AV1编码。此外,每个数据点还包含时间戳、帧索引、episode索引和任务索引等元数据。数据集适用于机器人模仿学习、策略学习等研究任务,数据以分块Parquet文件和MP4视频文件的形式组织。
This dataset is a robotic control dataset created using the LeRobot framework. It contains 271 training episodes, totaling 231,525 data frames and covers a single task. The data is stored in a structured format, including robotic arm action commands, state observations and visual observations. Specifically, both the action commands and state observations are 6-dimensional floating-point vectors, corresponding respectively to the six joint positions of the robotic arm: shoulder translation, shoulder lift, elbow flexion, wrist flexion, wrist rotation and gripper position. The visual observations are video streams from a front-facing camera, with a resolution of 1080x1920, a frame rate of 30fps, and encoded using AV1. Additionally, each data point also includes metadata such as timestamps, frame indices, episode indices and task indices. This dataset is suitable for research tasks such as robotic imitation learning and policy learning, and the data is organized into chunked Parquet files and MP4 video files.




