RoboCOIN/AI2_Alphabot_2_unplug_data_cable
收藏资源简介:
这是一个名为AI2_Alphabot_2_unplug_data_cable的机器人操作数据集。该数据集基于LeRobot格式进行了扩展,并完全兼容LeRobot。主要任务涉及使用名为AI2_Alphabot_2的双臂机器人,在家庭厨房场景中,用右手拿起白色的多端口USB集线器,同时用左手拔下白色的数据线。数据集包含932个episodes,总计371,680帧,帧率为30 FPS,数据大小为7.08 GB。机器人配置为双指末端执行器,通过VR控制器进行遥操作。传感器包括四个RGB摄像头:前胸摄像头、前头摄像头、左手腕摄像头和右手腕摄像头,每个摄像头的分辨率为640x480,视频编码为av1。数据集结构包括视频文件、状态数据(关节位置、速度等)、动作数据和元数据,组织在parquet文件中。此外,数据集提供了丰富的注释,如末端执行器加速度、方向、速度以及夹爪活动等,以支持多样化的学习方法。数据集由北京智源人工智能研究院(BAAI)的RoboCOIN团队贡献,遵循Apache 2.0许可证。
This is a robotic manipulation dataset named AI2_Alphabot_2_unplug_data_cable. It is extended based on the LeRobot format and fully compatible with it. The primary tasks involve using the dual-arm robot AI2_Alphabot_2 to grasp a white multi-port USB hub with its right arm, while unplugging a white data cable with its left arm in a home kitchen scenario. This dataset contains 932 episodes, totaling 371,680 frames at 30 FPS, with a total data size of 7.08 GB. The robot is equipped with two-fingered end-effectors and is teleoperated via VR controllers. The sensor suite includes four RGB cameras: a front-chest camera, a front-head camera, a left-wrist camera, and a right-wrist camera. Each camera has a resolution of 640×480, and the video is encoded in the AV1 format. The dataset structure consists of video files, state data (including joint positions, velocities, etc.), action data, and metadata, all organized in Parquet files. Additionally, the dataset provides rich annotations such as end-effector acceleration, orientation, velocity, and gripper activity to support diverse learning approaches. This dataset is contributed by the RoboCOIN team from the Beijing Academy of Artificial Intelligence (BAAI), and is released under the Apache 2.0 license.




