lavawolfiee/so101_bimanual_paper_fold_test
收藏资源简介:
该数据集是使用LeRobot创建的机器人数据集,专注于双手机器人(机器人类型为bi_so_follower)的演示数据。数据集包含3个episodes,总帧数为13279,涉及1个任务。数据以30fps的帧率采集,包括动作和观察特征:动作由12个浮点数值组成,表示左右机械臂的关节位置(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);观察包括状态(同样为12个关节位置)和来自多个视角的图像(如左上、左跟随者左、右跟随者右),图像分辨率为480x640,视频格式为mp4,无音频。数据集结构以parquet文件存储,适用于机器人学习任务,如模仿学习或强化学习,但具体任务(如paper_fold)未在描述中明确说明,仅从数据集名称推断可能与折纸相关。
This dataset was created using LeRobot and focuses on demonstration data for a bimanual robot (robot type bi_so_follower). It contains 3 episodes with a total of 13279 frames and involves 1 task. The data is collected at 30fps and includes action and observation features: actions consist of 12 float values representing joint positions for both left and right robot arms (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions); observations include state (similarly 12 joint positions) and images from multiple perspectives (e.g., left_top, left_follower_left, right_follower_right), with image resolution of 480x640 in mp4 video format without audio. The dataset structure is stored in parquet files and is suitable for robotics learning tasks such as imitation learning or reinforcement learning, though the specific task (e.g., paper_fold inferred from the dataset name) is not explicitly described in the README.




