Fine-Tuned Face Recognition Models for Sibling Discrimination

Face recognition has witnessed remarkable progress in recent years, with applications ranging from security systems to social media platforms. This work introduces a novel approach to the development of face recognition models specifically tailored for the challenging task of sibling discrimination. Recognising subtle differences among individuals who share familial resemblance poses a unique set of challenges for conventional face recognition systems. Four highly developed models used for face recognition, namely, VGG19, VGG16, VGGFace, and FaceNet, are fine-tuned with a Softmax activation function to classify images as same or different. Performance metrics derived from custom datasets show that the proposed approach yields robust and realistic results compared to pre-trained models, with varying accuracies for different facial parts.