Convolution Neural Network for Facial Kinship Verification

The process of detecting whether two people in a given pair of face pictures are biologically related or not is known as facial kinship verification. Deep learning-based techniques, in particular Convolutional Neural Networks (CNNs), have excelled at this challenge in recent years. In this article, we will compare a number of pre-trained CNN models and hybrid models to one another. Also, try to check performance by ensembling these models. The models are trained on the large-scale KinFaceW-I dataset and evaluated on the KinFaceW- II dataset, achieving state-of-the-art performance. In order to evaluate the performance, we have made our new dataset similar to the KinFaceW dataset. Additionally, our technique exhibits resilience to a variety of facial variables, including alterations in age, posture, and expression. Overall, can get a potential answer to the problem of facial kinship verification, which is crucial in numerous disciplines such as forensic investigation, family history research, and social media analysis. At last paper identified a ensembled or single models which work well on KinFace dataset and new dataset introduced by this paper.