chanq/record-sponge

该数据集使用LeRobot创建，专注于机器人领域，具体针对so_follower机器人类型。数据集包含32个训练片段，总计35664帧，覆盖1个任务。数据以30帧每秒的速率采集，存储为parquet和mp4格式，总数据文件大小为100MB，视频文件大小为200MB。特征包括动作（如肩部、肘部、手腕和夹爪的位置）、观测状态（与动作相同的位置信息）、手腕图像观测（480x640分辨率，3通道，30fps视频）、时间戳、帧索引、片段索引、索引和任务索引。数据集结构分为训练集，适用于机器人控制和模仿学习任务。This dataset was created using LeRobot and focuses on the robotics domain, specifically for the so_follower robot type. It includes 32 training episodes with a total of 35664 frames, covering 1 task. Data is collected at 30 frames per second, stored in parquet and mp4 formats, with a total data file size of 100MB and video file size of 200MB. Features include actions (e.g., positions of shoulder, elbow, wrist, and gripper), observation states (same position information as actions), wrist image observations (480x640 resolution, 3 channels, 30fps video), timestamps, frame indices, episode indices, indices, and task indices. The dataset structure is split into training sets, suitable for robot control and imitation learning tasks.