Aasdfip/atrium_grape_2

该数据集是一个机器人学习数据集，由LeRobot创建，专门用于双手机器人（bi_so_follower类型）的模仿学习或强化学习任务。数据集包含21个完整的情节（episodes），总计49189帧数据，覆盖4个不同的任务。数据以30帧每秒（fps）的速率采集，存储为Parquet格式文件，总数据文件大小为100MB，视频文件大小为200MB。数据集的特征包括：动作（action）和状态观测（observation.state），均为12维浮点向量，对应左右机械臂的12个关节位置（如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部滚动和夹爪位置）；图像观测（observation.images）来自三个摄像头视角：左腕摄像头、右腕摄像头和右顶部摄像头，每个视频帧分辨率为480x640像素，3通道彩色，使用AV1编解码器；此外，还包括时间戳（timestamp）、帧索引（frame_index）、情节索引（episode_index）、索引（index）和任务索引（task_index）等元数据。数据集仅包含训练分割（splits: train），适用于机器人控制、视觉导航等研究。This dataset is a robotics learning dataset created using LeRobot, specifically designed for imitation learning or reinforcement learning tasks with a bimanual robot (type: bi_so_follower). It contains 21 complete episodes, totaling 49189 frames, and covers 4 different tasks. Data is collected at 30 frames per second (fps) and stored in Parquet format files, with a total data file size of 100MB and video file size of 200MB. The dataset features include: action and observation.state, both as 12-dimensional float vectors corresponding to the positions of 12 joints on the left and right arms (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions); image observations (observation.images) from three camera perspectives: left wrist, right wrist, and right top, each video frame with a resolution of 480x640 pixels, 3 color channels, using AV1 codec; additionally, metadata such as timestamp, frame_index, episode_index, index, and task_index are included. The dataset only includes a training split (splits: train) and is suitable for research in robot control, visual navigation, etc.