KANFace

KANFace数据集是由帝国理工学院和雅典大学合作创建的，是目前最全面的大型手动标注面部图像和视频数据集。该数据集包含41,036张静态图像和1450万视频帧，涵盖了1,045个不同年龄（0至100岁）的个体。数据集详细标注了个体的身份、确切年龄、性别和亲缘关系，特别适用于年龄估计、性别识别和亲缘验证等任务。KANFace数据集的创建旨在解决深度学习模型在面部分析中存在的偏见问题，通过提供多样化的数据来提高模型的泛化能力和公平性。

The KANFace Dataset was collaboratively developed by Imperial College London and the National and Kapodistrian University of Athens, and is currently the most comprehensive large-scale manually annotated facial image and video dataset. It contains 41,036 static images and 14.5 million video frames, covering 1,045 individuals with ages ranging from 0 to 100 years old. Detailed annotations are provided for individual identity, exact age, gender and kinship relationship, making it particularly suitable for tasks such as age estimation, gender recognition and kinship verification. The KANFace Dataset was created to address the bias issue in deep learning models for facial analysis, aiming to improve the generalization ability and fairness of models by supplying diverse training data.