HARPER
收藏资源简介:
HARPER数据集由意大利维罗纳大学与英国格拉斯哥大学联合创建,专注于机器人视角下的三维人体姿态估计与预测。该数据集以波士顿动力公司的四足机器人Spot为研究对象,由于机器人的视角接近地面,只能捕获到人体的一部分,这为三维姿态的解析带来了不小的挑战。HARPER 数据集涵盖了 15 种不同的互动动作,其中 10 种涉及到机器人与用户的直接物理接触,数据集中不仅包括 Spot 机器人内置的立体摄像头所拍摄的视频,还包含了由六摄像头 OptiTrack 系统同步录制的视频资料,确保了骨骼模型的精确度达到毫米级以下。该数据集的应用领域广泛,特别是在人机协作、行为识别、意图预测等方面,为研究者提供了一个宝贵的资源。
The HARPER dataset was jointly created by the University of Verona in Italy and the University of Glasgow in the UK, focusing on 3D human pose estimation and prediction from a robotic perspective. It takes Boston Dynamics' quadruped robot Spot as its research subject. As the robot's viewpoint is close to the ground, it can only capture partial views of the human body, which poses considerable challenges to 3D pose analysis. The HARPER dataset covers 15 distinct interactive actions, 10 of which involve direct physical contact between the robot and human users. The dataset not only includes videos captured by the built-in stereo camera of the Spot robot, but also video data synchronously recorded by a six-camera OptiTrack system, ensuring that the skeletal model achieves a sub-millimeter level of accuracy. This dataset has a wide range of application scenarios, especially in human-robot collaboration, action recognition, intent prediction and other fields, providing a valuable resource for researchers.




