FaceSkin
收藏资源简介:
FaceSkin是由招商银行AI实验室创建的大规模人脸皮肤斑点数据集,旨在为深度学习提供隐私保护的分类任务数据。该数据集包含多样化的年龄和种族信息,并引入了合成皮肤斑点,来源于2D和3D攻击图像,如打印纸、重放和3D面具,以扩展应用场景。数据集的创建过程中,使用了Mediapipe进行面部标志定位,提取纯皮肤斑点,并通过DeepFace进行属性标注,如年龄、种族和性别。FaceSkin的应用领域包括人脸反欺诈和年龄估计,旨在解决人脸图像隐私保护和属性分类的需求。
FaceSkin is a large-scale facial skin lesion dataset created by the AI Lab of China Merchants Bank, aiming to provide privacy-preserving classification task data for deep learning. This dataset contains diverse age and ethnicity information, and introduces synthetic skin lesions sourced from 2D and 3D attack images such as printed paper, replay attacks and 3D masks to expand application scenarios. During the dataset construction process, Mediapipe was used for facial landmark localization to extract pure skin lesions, and attributes including age, ethnicity and gender were annotated via DeepFace. The application fields of FaceSkin cover face anti-fraud and age estimation, which is designed to address the requirements of facial image privacy protection and attribute classification.



