This paper evaluated the impact of synthetic images on Morphing Attack
Detection (MAD) using a Siamese network with a semi-hard-loss function. Intra
and cross-dataset evaluations were performed to measure synthetic image
generalisation capabilities using a cross-dataset for evaluation. Three
different pre-trained networks were used as feature extractors from traditional
MobileNetV2, MobileNetV3 and EfficientNetB0. Our results show that MAD trained
on EfficientNetB0 from FERET, FRGCv2, and FRLL can reach a lower error rate in
comparison with SOTA. Conversely, worse performances were reached when the
system was trained only with synthetic images. A mixed approach (synthetic +
digital) database may help to improve MAD and reduce the error rate. This fact
shows that we still need to keep going with our efforts to include synthetic
images in the training process.

研究评估了使用具有半硬损失函数的暹罗网络在形态攻击检测（MAD）中，合成图像对其影响。通过跨数据集评估来测量合成图像的泛化能力。使用传统的 MobileNetV2、MobileNetV3 和 EfficientNetB0 预训练网络作为特征提取器。我们的结果表明，在 FERET、FRGCv2 和 FRLL 数据库上， EfficientNetB0 训练的 MAD 具有比 SOTA 更低的错误率。相反，当系统仅使用合成图像进行训练时，性能较差。混合方法（合成 + 数字）数据库可以帮助改进 MAD 并降低错误率。这表明我们仍需要在训练过程中继续努力包含合成图像。