junpark-ai/jh-buckle

该数据集是一个机器人学相关数据集，专门用于ffw_bg2_rev4_arms_only机器人类型。它包含50个训练集，总计8945帧，覆盖1个任务，并包含150个视频。数据以30帧每秒的帧率采集，分割为训练集（0:50）。数据集结构包括时间戳、帧索引、集索引、索引、任务索引等特征，以及来自三个摄像头的观察图像（头摄像头分辨率376x672，左腕摄像头分辨率240x424，右腕摄像头分辨率240x424），所有图像均为彩色（3通道）。此外，还包括16维的观察状态（如左右臂和抓取器的关节位置）和16维的动作数据（对应关节控制）。数据以Parquet格式存储，视频以MP4格式存储，使用libx264编解码器。数据集使用Apache 2.0许可证，并基于Physical AI Tools和LeRobot工具创建。This is a robotics-specific dataset dedicated to the ffw_bg2_rev4_arms_only robot type. It contains 50 training sets, totaling 8945 frames, covering 1 single task, and includes 150 videos. All data was collected at a frame rate of 30 frames per second (FPS), and the dataset is split into the training set (0:50). The dataset structure includes features such as timestamp, frame index, set index, index, task index, etc., as well as observation images from three cameras: the head camera with a resolution of 376x672, the left wrist camera with a resolution of 240x424, and the right wrist camera with a resolution of 240x424. All images are 3-channel color images. Additionally, it includes 16-dimensional observation states (such as joint positions of the left and right arms and grippers) and 16-dimensional action data corresponding to joint control. The data is stored in Parquet format, while the videos are stored in MP4 format using the libx264 codec. This dataset is licensed under the Apache 2.0 license, and was created based on the Physical AI Tools and LeRobot toolkits.