HUP-3D
收藏资源简介:
HUP-3D是由Wellcome/EPSRC Centre for Interventional and Surgical Sciences (WEISS), University College London创建的一个3D多视角合成数据集,专门用于妇产科超声中的手-超声探头姿态估计。该数据集包含超过31,680组RGB、深度和分割掩码帧,以及相关的姿态真实数据,强调图像的多样性和复杂性。数据集的创建采用了基于相机视角的球体概念,通过预训练网络生成多种手抓握姿态,并结合软件渲染技术,增加了手和手臂纹理、光照条件和背景图像的多样性。HUP-3D数据集主要应用于混合现实医学教育领域,旨在通过分析手和探头的运动,提供定制化的指导和辅导应用,解决当前缺乏有效超声探头指导标准的问题。
HUP-3D is a 3D multi-view synthetic dataset created by the Wellcome/EPSRC Centre for Interventional and Surgical Sciences (WEISS), University College London, specifically designed for hand-ultrasound probe pose estimation in obstetrics and gynecology ultrasound. This dataset contains over 31,680 sets of RGB, depth and segmentation mask frames along with corresponding ground-truth pose data, emphasizing the diversity and complexity of the images. The dataset was developed using the concept of a camera-view-based sphere, where pre-trained networks are used to generate diverse hand grasping poses, combined with software rendering techniques to increase the diversity of hand and arm textures, lighting conditions and background images. The HUP-3D dataset is primarily applied in the field of mixed reality medical education, aiming to provide customized guidance and tutoring applications by analyzing the movements of the hand and the ultrasound probe, addressing the current lack of effective standards for ultrasound probe guidance.




