Kinship Verification Based on Various Models (PATCH, STATIC, SVM, CNN, LPQ, BSIF,LBP)

Abstract

Kinship verification from facial images has emerged as a significant area of research in computer vision, with growing interest due to its wide range of applications. Automatically determining whether two individuals share a biological relationship based solely on their facial features holds potential in fields such as family tree reconstruction, organizing family photo albums, annotating images, locating missing persons, and forensic investigations. This project aims to design and implement a robust artificial intelligence model for kinship verification by leveraging biometric traits, particularly facial features. Utilizing advancements in machine learning, pattern recognition, and statistical analysis, the goal is to accurately determine familial relationships between individuals. The key stages of the project include data collection, preprocessing, feature extraction, model training, and evaluation. Our approach focuses on the KIN Face II dataset, employing advanced preprocessing techniques to enhance image quality, followed by deep CNN-based feature extraction to identify kinship-related patterns. The final model is trained to learn discriminative features that effectively distinguish kin from non-kin. The expected outcome of this work is a reliable and high-performing model for kinship verification, contributing to the broader scope of biometric identification systems and offering practical solutions for various real-world applications.