LAUSR

Abstract Kinship verification is an emerging task in computer vision which aims at finding out whether there is a kin relation between given identities through their facial images. Applications of kinship verification include image annotation, children adoption, and social media analysis, etc. However, kinship verification through facial images is challenging because facial images usually contain high intra-variances, which vary from genetic, age and gender difference. Over the past decade, more and more effective methods have emerged. This paper aims at categorizing and evaluating these methods systematically. We attach great importance to the difficulties in practical applications of kinship verification, and review the prominent algorithms from the perspective of learning more efficient models with more diverse kin relations. Then we further show how to develop an efficient and robust kinship verification system. Finally we present several potential directions for future research.