AN EFFICIENT KINSHIP VERIFICATION AND FAMILY TREE CONSTRUCTION FRAMEWORK USING GRAPHICAL LEARNING AND KNOWLEDGE REPRESENTATION

Abstract

Kinship verification and identification from image data has been a challenging and complex field in computer vision over the last two decades. It determines the blood relation between two people from their facial images. It has numerous applications, including finding missing children, forensic research, and is also used in genealogical research and the reconstruction of family history. This paper presents a novel approach for automatically creating family trees from a set of family pictures up to three levels. Traditional methods rely on manual research and rule-based systems, which can be time-consuming and prone to errors. The proposed approach is a hybrid solution that combines the strengths of graphical neural networks with the accuracy of mathematical logic. First, the GCN technique was applied to learn an underlying representation of individuals from their facial features and identify undirected kin relations. These relationships are then expressed using first-order predicate logic to determine directionality and construct a comprehensive family tree. Creating a family tree plays a vital role in establishing individual identity by linking each person to their family members. This effort directly supports Sustainable Development Goal (SDG) 16.9, which aims to provide legal identity for all. We demonstrate our approach on the FIW dataset, and the results show that our approach works efficiently and has improved accuracy compared to traditional methods.