3D Mannequin Dataset for Face Anti-Spoofing

The 3D Mannequin Dataset for Face Anti-Spoofing is a specialized liveness-detection collection designed to challenge face recognition systems with realistic non-human targets. It features over 1,000 images of retail mannequins—including male, female, and child figures, some styled with natural hair—captured from both selfie and frontal viewpoints to simulate real-world presentation attacks. The dataset includes mannequins accessorized with glasses, sunglasses, scarves, and hats, further increasing diversity and realism. By using 3D textures and lifelike appearances, this dataset helps researchers and developers enhance anti-spoofing and liveness-detection models, training them to distinguish human faces from highly deceptive mannequin faces. Hosted on Kaggle, it is accessible under a compact size (about 3 MB in 20 files) and has been downloaded approximately 146 times, reflecting its growing use in the community.

面向人脸防欺骗（Face Anti-Spoofing）的3D模特数据集是一款专为活体检测（liveness-detection）打造的专属数据集，旨在通过逼真的非人类目标对人脸识别系统进行性能测试。该数据集包含超1000张零售模特图像，涵盖男性、女性及儿童模特，部分模特配有仿真发丝，采集视角包含自拍视角与正脸视角，以模拟真实场景下的呈现式攻击（presentation attacks）。数据集内的模特还搭配了眼镜、太阳镜、围巾与帽子等配饰，进一步提升了数据集的多样性与真实感。该数据集采用3D纹理与逼真外观设计，可帮助研究人员与开发者优化防欺骗与活体检测模型，训练模型区分真人面孔与极具欺骗性的模特面孔。该数据集托管于Kaggle平台，体积小巧（20个文件总计约3MB），目前累计下载量约146次，体现出其在社区中的使用率正逐步提升。