A HIGHLY EFFICIENT FACIAL RECOGNITION FOR BIOMETRIC APPLICATIONS USING CONVOLUTIONAL NEURAL NETWORK

Abstract: The most common feature that people use to identify one other is their faces. In computer vision and image understanding, face recognition (FR) is a traditional problem that is still actively pursued. This function of biometrics, which is a component of deep learning, has long been met. Face recognition has grown in importance as a tool for safety and security in recent years. Measuring and examining each person's distinct physical or behavioral traits is the aim of biometric systems. Due to its ability to gather biometric data without invading personal space or requiring the agreement of the user, facial biometrics have become one of the most popular biometric data types in recent years. Face recognition has seen a recent surge in interest in deep learning, particularly in deep convolutional neural networks, and several deep learning techniques have been put forth. In the proposed work, a comparison is made between the holistic and fisher face approaches and the face recognition application using Convolutional Neural Networks (CNN) using Python. In contrast to traditional techniques, the suggested methodology normalizes the probability distribution by employing pooling layers, extra convolutional layers with ReLU layers, a fully connected layer, and a Softmax Loss Layer. 1500 photos with various facial expressions make up the dataset, and the CNN method is used to train and evaluate the model to achieve accuracy. Based on experimental data, the suggested Neural Network received a score of 98.96% accuracy.

Keywords: Facial Recognition, biometrics, deep Learning, convolutional neural networks, facial dataset, data security.