Thesis (MSc) - Cyprus International University. Institute of Graduate Studies and Research Information Systems Engineering

Due to the recent outbreak of the COVID-19 virus, the need for an online management system became needful and with many studies already carried out, this study focuses on the evaluation of the algorithms available currently. An extensive analysis of face detection and recognition models within the framework of proctoring systems is presented in this thesis. The research focuses on face detection using the Single Shot MultiBox Detector (SSD) which is a Convolutional Neural Networks (CNN) model, in conjunction with Convolutional Neural Networks (CNN), Prototypical Networks, and Capsule Neural Networks for face recognition. Key parameters like accuracy, precision, recall, and computing time are included in the evaluation, which is conducted using a custom dataset for testing and evaluation, and the training datasets Labelled Faces in the Wild (LFW) and Celebrities in Frontal Profile in the Wild (CFPW). The results show that CNN consistently performs well, which makes it a strong contender for proctoring systems. Capsule Networks perform well whereas Prototypical Networks perform well when dataset limitations are met. In a variety of situations, the SSD model demonstrates proficiency in face detection tasks. The paper acknowledges the limits of the dataset and the significance of adding several points of view and suggests that while specific needs determine which model is best, CNN is the most popular face recognition method for a proctoring system. According to the consequences, proctoring systems may be made much more reliable and effective by integrating these methods, which will help provide a secure method for monitoring exams.