Webface-OCC
收藏资源简介:
Webface-OCC是由多媒体软件国家工程研究中心和武汉大学计算机学院创建的一个大型人脸识别数据集,专注于解决人脸遮挡问题。该数据集包含804,704张人脸图像,涵盖10,575个不同个体,通过模拟多种遮挡类型(如眼镜和口罩)来增强数据集的多样性和真实性。创建过程中,研究团队首先收集了多种遮挡物,并通过随机组合其属性(如物体、纹理和颜色)来生成大量真实遮挡类型。数据集的应用领域主要集中在提高模型在实际场景中对遮挡人脸的识别能力,特别是在COVID-19疫情期间,口罩遮挡对人脸识别技术提出了新的挑战。
Webface-OCC is a large-scale face recognition dataset developed by the National Engineering Research Center for Multimedia Software and the School of Computer Science, Wuhan University, which focuses on addressing the problem of face occlusion. This dataset contains 804,704 face images covering 10,575 distinct individuals, and enhances its diversity and realism by simulating various occlusion types such as glasses and face masks. During the dataset construction, the research team first collected a variety of occluding objects, then generated a large number of realistic occlusion types by randomly combining their attributes including object category, texture and color. The dataset is mainly applied to improving the recognition performance of models for occluded faces in real-world scenarios, especially during the COVID-19 pandemic, when face mask occlusion posed new challenges to face recognition technologies.

- 1When Face Recognition Meets Occlusion: A New Benchmark多媒体软件国家工程研究中心,武汉大学计算机学院 · 2021年



