Siamese-Based Face Recognition using Convolutional Neural Networks

: Given its potential applications in fields ranging from security and healthcare to biometrics and marketing, face recognition has emerged as an important area of research. Therefore, this technique has received extensive attention due to its scalability, high recognition rate, and uniqueness of face. Therefore, deep learning, particularly Convolutional Neural Networks (CNNs), became a very changed method in recent years. Siamese Neural Networks is a type of architecture used for deep learning, which is well-suited for problems where we compare and rank images, so the face recognition tasks. We trained the model, which was built with a CNN, on three types of data: positive and anchor images captured through the webcam with OpenCV and negative samples pulled from the Labeled Faces in the Wild dataset. Not only was the performance on the Test set very good (accuracy: 97%; precision: 96%; recall: 98%; F1 score: 96%) these metrics also illustrate a great performance on the model to differentiate between the matching and quickly not comparable faces with confidence. Also, its lightweight and memory-efficient system can achieve fast and accurate face detection and identification using OpenCV, which is practical for real applications.