o-hisa/eval_pick_lego_in_several_positions

该数据集是通过LeRobot创建的，专注于机器人抓取乐高积木在不同位置的表现。数据集包含16个episodes，共15984帧，记录了机器人手臂的动作（如肩部、肘部、腕部的位置）、观察状态以及来自顶部、腕部和前部摄像头的视频数据。数据以parquet格式存储，视频数据以mp4格式存储。数据集还包含任务索引、时间戳等元数据。评估报告显示，模型在已学习的位置上表现良好，但在未学习的位置上存在抓取失败的情况，并提出了增加学习位置和改进抓取策略的建议。

This dataset was created using LeRobot and focuses on the performance of a robotic arm in picking Lego blocks in various positions. It includes 16 episodes with a total of 15,984 frames, recording the robots actions (e.g., positions of shoulder, elbow, wrist), observation states, and video data from top, wrist, and front cameras. The data is stored in parquet format, and video data is stored in mp4 format. The dataset also includes metadata such as task indices and timestamps. The evaluation report indicates that the model performs well in learned positions but struggles in unlearned positions, with suggestions for increasing learning positions and improving grasping strategies.