Machine Learning (ML) has shown progressive growth in recent years, with deep learning (DL) emerging as an effective method for analyzing complicated data and generating accurate inferences. Diverse fields like computer vision, pattern recognition and natural language processing extensively utilize DL models, drawing motivation from the structure and function of the human brain. DL has confirmed significant promise in the field of biometric systems, which depend on distinct physiological and behavioral characteristics for identification and verification. This study provides a comprehensive assessment of several DL methods aimed at enhancing biometric systems, specifically in the field of face-recognition. This research suggests a methodology for evaluating DL models such as AlexNet, VGGFace, FaceNet, and ResNet in order to identify the best appropriate model for facial recognition applications. The purpose of this framework is to guide the development of biometric systems that are both more precise and efficient. We conducted the assessment on a real facial dataset that we collected from Kaggle. We use accuracy, precision, recall, and F1-measure parameters to compare the performance of different models. In ResNet, the accuracy is 98%, which is higher than the other three models.