Enhancing Classroom Attendance Systems with Face Recognition through CCTV using Deep Learning

This paper investigates the effectiveness of face recognition-based attendance systems, focusing on deep learning models applied to automate attendance tracking in classroom settings. Attendance management is critical for educational institutions, but traditional methods are often plagued by inefficiencies and inaccuracies. To address these issues, we explored the use of advanced face recognition technology, particularly leveraging CCTV-based systems combined with deep learning models namely ResNet-50 and VGG-16. We used images collected from actual classroom environments, applied image dataset augmentation methods, used transfer learning for features extraction and Principal Component Analysis (PCA) for dimensionality reduction and then used the data as an input in the fine-tuned pre-trained models. Our experimental analysis using PCA as a feature extractor indicate that the ResNet-50 model achieved precision and recall rate of 96% and 93% respectively and VGG16 achieved precision and recall rate of 92% and 89% respectively. Despite promising results, challenges remain, such as the impact of camera quality and inherent limitations such as viewing geometry in current face recognition technologies. Future improvements may involve upgrading camera systems and employing image enhancement techniques. This study offers insights into the practical application of face recognition for attendance and suggests directions for enhancing system reliability and accuracy.