Purpose: Face recognition has become an increasingly intriguing field in artificial intelligence research. In this study,   This study aims to explore the application of CNNs, implemented through TensorFlow, to develop a robust model for enhancing facial recognition accuracy in student attendance systems. The focus of this research is the development of a model capable of recognizing student faces under various lighting conditions and poses in an academic environment, using a multi-class dataset of student images collected from internship attendance records at the Computer Science Department.
Methods: The dataset, comprising facial images from 19 students, served as the foundation for training and validating the CNN model. The dataset, sourced from the computer science department's internship attendance records, included a total of 231 images for training and 59 images for validation. The preprocessing phase included facial area detection and categorization, resulting in a well-organized dataset for training and validation. The CNN architecture, consisting of seven layers, was meticulously designed to achieve optimal performance.
Result: The model exhibited exceptional accuracy, reaching 93% on the validation dataset after 300 training epochs. Precision, recall, and F1-score metrics were employed for a detailed evaluation across diverse classes, highlighting the model's proficiency in accurately categorizing facial images. Comparative analyses with a VGG-16-based model showcased the superiority of the proposed CNN architecture. Moreover, the implementation of a web service demonstrated the practical applicability of the model, providing accurate predictions with a remarkable response time of less than 0.3 seconds.
Novelty: This comprehensive study not only advances face recognition technology but also presents a model applicable to real-world scenarios, particularly in student attendance systems.