FACE RECOGNITION USING HYBRID QUANTUM CONVOLUTIONAL NEURAL NETWORK (QCNN) AND IMPROVED FEATURE EXTRACTION TECHNIQUES

Abstract

Face recognition is an advanced biometric technology that focuses on accurately identifying facial features to differentiate individuals, even when faces appear highly similar or identical. Modern face recognition systems are much more dependable and effective at managing difficult obstacles because of rapid growth of computer vision and deep learning, incorporating variations in lighting, position, facial expressions, and occlusions. Analysing minute variations in facial traits that may not be easily discernible to the human eye is the main objective of same face identification. Additionally, since the human face is a three-dimensional (3-D) entity, it may be illuminated unevenly and with a distorted perspective. Consequently, it may not be possible to identify a real face. This research work introduces a hybrid automatic detection and intelligent feature extraction model for face recognition that leverages two-dimensional (2D) facial images sourced from diverse origins to construct three-dimensional (3D) face meshes using 468 MediaPipe landmarks. Preprocessing pipeline includes use of Restormer to improve image quality, then use of segmentation based on deep learning algorithms like Mask R-CNN to isolate the foreground (face) from the background to decrease noise and increase detection accuracy. Extraction of features is done by use of Vision Transformer (ViT). Features are then classified through hybrid deep classifiers in the form of FaceNET (FN) with quantum convolutional neural network (QCNN). This is to ensure accuracy despite the presence of different facial poses and expressions.