liveness-detection-dataset
收藏资源简介:
该数据集是一个全面的面部活体检测数据集,专为面部反欺骗、生物识别面部识别和呈现攻击检测(PAD)系统设计。与仅覆盖一两种攻击类型的公共基准不同,该数据集集成了所有主要的呈现攻击类别,包括纸质攻击、重放攻击和3D面具攻击(硅胶、乳胶、纸包、树脂、布料),提供了超过10万条视频,涵盖11种以上标记的攻击类型,并映射到iBeta Level 1、Level 2和Level 3 PAD认证。数据集包含真实面部(bona fide)和多种攻击类型的视频序列,每种攻击类型均标有其iBeta级别。数据采集使用了现代智能手机(如iPhone 14/13 Pro、Galaxy S23、Pixel 7等),覆盖室内外环境、多种光照条件和平衡的人口统计特征(包括高加索人、黑人、亚洲人和拉丁裔)。适用于训练二元反欺骗分类器、多类攻击类型分类器、iBeta认证准备以及跨攻击泛化研究。
This is a comprehensive face liveness detection dataset designed for face anti-spoofing, biometric facial recognition, and presentation attack detection (PAD) systems. Unlike public benchmarks that only cover one or two attack types, this dataset integrates all major presentation attack categories, including paper attacks, replay attacks, and 3D mask attacks (silicone, latex, paper-wrapped, resin, fabric). It contains over 100,000 video sequences covering more than 11 labeled attack types, and maps to iBeta Level 1, Level 2 and Level 3 PAD certifications. The dataset includes bona fide facial video sequences and video sequences of various attack types, with each attack type labeled with its corresponding iBeta level. Data was collected using modern smartphones (e.g., iPhone 14/13 Pro, Galaxy S23, Pixel 7, etc.), covering indoor and outdoor environments, diverse lighting conditions, and balanced demographic characteristics including individuals from Caucasian, Black, Asian and Latino backgrounds. It is suitable for training binary anti-spoofing classifiers, multi-class attack type classifiers, iBeta certification preparation, and cross-attack generalization research.




