rollout_openarm-inference2_20260724_194403

该数据集由LeRobot创建，是一个机器人技术领域的数据集，专为双臂机器人（bi_openarm_follower）设计。它包含一个总时长为208帧的单个训练集，帧率为30fps，以parquet文件格式存储，总数据文件大小为100MB，视频文件大小为200MB。数据集特征包括：动作数据（action），包含左右双臂各7个关节位置和左右夹爪位置，共16个浮点数值；观测状态（observation.state），与动作数据类似，包含16个关节位置；观测图像，包括左腕摄像头（left_wrist）、基座摄像头（base）和右腕摄像头（right_wrist）的视频数据，每个视频分辨率为480x640，3通道彩色，使用av1编解码器；以及时间戳、帧索引、片段索引、索引和任务索引等元数据。数据集结构支持训练任务，并包含视频文件路径信息。This dataset was developed using LeRobot, and is a robotics-focused dataset specifically designed for the dual-arm robot (bi_openarm_follower). It includes a single training split with a total of 208 frames at a frame rate of 30 fps. The data is stored in Parquet file format, with a total size of 100 MB for the core data files and 200 MB for the video files. The dataset features are as follows: 1. Action data (action): contains 7 joint positions for each of the left and right arms, plus the positions of the left and right grippers, totaling 16 floating-point values; 2. Observation state (observation.state): similar to the action data, containing 16 joint positions; 3. Observation images: video data from the left wrist camera (left_wrist), base camera (base), and right wrist camera (right_wrist). Each video has a resolution of 480x640, is 3-channel color, and uses the AV1 codec; Additionally, metadata such as timestamps, frame indices, episode indices, indexes, and task indices are included. The dataset structure supports training tasks and contains video file path information.