Deep Learning Model for Facial Kinship Verification Using Childhood Images*

Abstract

Kinship Verification (KV) using facial images is a challenging research problem in computer vision field. It is a binary classification problem. Mainly, this research works on similarities of two facial images. If the age gap is more between parent's image and their children then their facial similarities are dispersed. Further, it is quite challenging task to assess their kinship. This is called age based kinship verification problem and it is one of the research gap. In this article, we proposed a model with childhood image database. This childhood database consists of parent and their children image considered at their childhood stage which is minimum age gap so that their facial similarities are more. Further, proposed a model with inception neural network to extract the facial features. In classification stage multiple dense layer are designed to assess the kinship of given test pairs from trained weights (W). Our proposed model experimented and achieved 92% accuracy for the proposed databases and also performed for other existed database in the wild which has shown better accuracy as compared with state of art methods.