ABSTRACT
Kinship verification is an active area of interest in research fields such as anthropology, neuroscience, psychology, and computer-vision. In this chapter, we focus on using computer vision techniques to determine if two persons are lineally related by analyzing their face images. Establishing such relationships have wide-ranging applications such as tagging family photo albums and locating missing relatives. This chapter describes kinship verification using two deep learning algorithms: stacked denoising autoencoders and deep belief networks. These kinship-verification algorithms yield state-of-the-art performance on five publically available kinship-face databases. We also demonstrate the efficacy of these approaches for the task of self-kinship or age-invariant face recognition on FG-NET and UB KinFace databases. These results illustrate the effectiveness of representation learning for the challenging problem of kinship verification.
