SymFace

SymFace数据集由挪威科技大学挪威生物识别实验室创建，包含77,034个样本，其中25,919个为合成身份。该数据集通过StyleGAN生成，旨在模拟真实世界中的面部变化，用于解决面部识别中的隐私和数据不平衡问题。创建过程中，利用PCA定义潜在空间中的语义方向，并通过预训练的面部识别系统控制原始样本与合成样本之间的相似性。SymFace数据集主要应用于面部识别性能测试，特别是在边境控制等高安全性应用中。

SymFace dataset was created by the Norwegian Biometric Laboratory at the Norwegian University of Science and Technology. It contains 77,034 samples, among which 25,919 are synthetic identities. Generated via StyleGAN, this dataset aims to simulate real-world facial variations and address the issues of privacy and data imbalance in facial recognition. During its development, PCA was used to define semantic directions in the latent space, and a pre-trained facial recognition system was employed to control the similarity between original and synthetic samples. The SymFace dataset is primarily applied for facial recognition performance testing, particularly in high-security applications such as border control.