JakeFDB, DebbieFDB, HarryFDB

本研究贡献了三个面部图像数据集：JakeFDB、DebbieFDB和HarryFDB，用于评估自适应异常检测框架。这些数据集包含年龄差异超过十年的个体面部样本，以纳入更长的时间背景。数据集通过从电视剧和电影中收集样本创建，每个数据集涵盖了从儿童到成年的面部变化。JakeFDB来自电视剧《Two and a Half Men》，DebbieFDB来自《Shameless》，HarryFDB则来自《Harry Potter》系列电影。这些数据集特别适合分析概念漂移，尤其是内部类和跨类漂移，为智能设备中的多面认证提供了重要的研究资源。

This study contributes three facial image datasets: JakeFDB, DebbieFDB, and HarryFDB, intended for evaluating adaptive anomaly detection frameworks. These datasets include facial samples of individuals with an age gap of over ten years to accommodate a longer temporal context. The datasets are constructed by collecting samples from television series and feature films, with each dataset covering facial changes from childhood to adulthood. Specifically, JakeFDB is sourced from the TV series *Two and a Half Men*, DebbieFDB from *Shameless*, and HarryFDB from the *Harry Potter* film franchise. These datasets are particularly well-suited for analyzing concept drift, especially intra-class and inter-class concept drift, and serve as a critical research resource for multi-face authentication in smart devices.