tatsuyaaaaaaa/so101_pick_red_dice_sim

该数据集是一个用于机器人学习的模拟数据集，专注于拾取红色骰子的任务。数据集包含75个训练集，总计62239帧数据，帧率为30fps。使用so_follower机器人类型，动作空间为6自由度关节位置控制（包括肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）。观测状态包括相同的关节位置，以及来自三个摄像头的视觉输入，每个摄像头提供480x640分辨率的RGB视频（编码为av1，无音频）。数据以parquet格式存储，视频以mp4格式存储。数据集基于LeRobot框架创建，适用于机器人控制、强化学习或模仿学习任务。

This dataset is a simulation dataset for robotics learning, focusing on the task of picking up a red dice. It contains 75 training episodes with a total of 62,239 frames at 30 fps. The robot type is so_follower, with an action space of 6-degree-of-freedom joint position control (including shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions). Observation states include the same joint positions and visual inputs from three cameras, each providing RGB video at 480x640 resolution (encoded in av1, without audio). Data is stored in parquet format, and videos are stored in mp4 format. The dataset was created using the LeRobot framework and is suitable for robotics control, reinforcement learning, or imitation learning tasks.