A robust kinship verification scheme using face age transformation

Kinship verification determines the existence of a kin relationship between people through facial image analysis, and can be used in various real-life applications, such as finding missing family members, analyzing social media, and genealogy research. Recently, many convolutional neural network (CNN)-based kinship verification methods have been proposed, owing to the good image-processing performance of the CNNs. Nevertheless, insufficient labeled data and age differences in the kinship images, make kinship verification quite difficult. To mitigate these limitations, herein, we first propose a face age transformation model to generate facial images of various age groups. Then, we construct a cross-age kinship verification model constructed using the generated images as a training dataset. To show the effectiveness of the proposed scheme, we conducted various comparative experiments with other models using popular kinship datasets and confirmed that our proposed method exhibited an improved verification accuracy.