UMDFaces
收藏资源简介:
UMDFaces是由马里兰大学帕克分校创建的一个大型人脸数据集,包含367,888个标注的人脸图像,涉及8,277个不同个体。该数据集通过人工标注和深度网络技术收集和标注,提供了高质量的人脸边界框、估计的姿态(滚动、俯仰和偏航)、21个关键点位置以及性别信息。此外,数据集中的关键点标注质量已通过人工验证。UMDFaces数据集适用于人脸识别、头部姿态估计和关键点定位等领域的研究,旨在推动人脸分析技术的进步和应用。
UMDFaces is a large-scale face dataset developed by the University of Maryland, College Park, which contains 367,888 annotated face images from 8,277 distinct individuals. Collected and annotated via both manual labeling and deep network technologies, this dataset provides high-quality face bounding boxes, estimated poses (roll, pitch, and yaw), 21 key point locations, as well as gender information. Additionally, the quality of the key point annotations in this dataset has been manually verified. The UMDFaces dataset is applicable to research in fields including face recognition, head pose estimation, and key point localization, with the goal of advancing the development and applications of face analysis technologies.




