Biological relation recognition from facial features

• Main Author: Lim, Iris Xin Yi
• Other Authors: Anamitra Makur
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/177033
• Institution: Nanyang Technological University

Facial recognition technology has seen remarkable advancements in recent years. One emerging area of interest within this domain is Facial Kinship Verification (FKV), which aims to identify biological relationships between individuals based on their facial features. Drawing from psychological insights suggesting that closely related individuals often exhibit visible facial similarities, this study explores the integration of deep learning techniques to discern such resemblances. Unlike conventional facial recognition tasks, FKV presents a distinctive challenge as it involves comparing facial similarities between pairs of individuals rather than identifying individuals in isolation. To address this challenge, this study proposes the use of Siamese Neural Networks, a specialized architecture adept at measuring similarities between pairs of images. In this study, we combine Siamese Neural Networks with transfer learning, a powerful technique that allows for the adaptation of pre-trained face recognition models to similar tasks like FKV. By leveraging pre-trained network on facial recognition tasks, the proposed approach aims to enhance the feature extraction capabilities of the Siamese network, thereby improving its performance in identifying familial relationships.