fhliang/so101_pour_train_4x_scene
收藏资源简介:
该数据集使用LeRobot创建,专注于机器人倒水任务,适用于模仿学习和视觉语言动作微调。数据集包含225个episodes,总计148,530帧,采样率为30 fps。机器人类型为so_follower,数据特征包括动作(如肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部滚动和夹爪位置)和观测状态(相同关节位置),以及场景图像观测(480x640分辨率,RGB三通道)。此外,还包含时间戳、帧索引、episode索引等元数据。数据以Parquet格式存储,视频以MP4格式提供,总数据大小约为100 MB,视频大小约为200 MB。
This dataset was created using LeRobot and focuses on a robot pouring task, suitable for imitation learning and visual-language-action fine-tuning. It includes 225 episodes with a total of 148,530 frames at 30 fps. The robot type is so_follower, and features encompass actions (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, gripper positions) and observation states (same joint positions), along with scene image observations (480x640 resolution, RGB three channels). Additionally, metadata such as timestamps, frame indices, episode indices are included. Data is stored in Parquet format, videos in MP4 format, with total data size around 100 MB and video size around 200 MB.




