DEEP LEARNING-POWERED ATTENDANCE TRACKING: A CONTACTLESS AND EFFICIENT LOGGING SYSTEM

Abstract

This paper proposes a contact less attendance monitoring system based on deep learning that seeks to increase efficiency, cleanliness, and accuracy in educational establishments. Manual roll calls or biometric fingerprint scanners considered a traditional attendance system are subject to manipulation, ineffective, and prone to hygiene issues and introduction in post-pandemic settings. This system takes advantages of Convolutional Neural Networks (CNNs) to bring a new real-time facial recognition model that is built with an intuitive interface. The system is built with Python, OpenCV, and TensorFlow, it identifies and confirms the identities of other students in real-time camera data and records their presence into a local database that has high security. To make the model accurate across different lighting, facial orientations and expression, the model was trained on a wide range of facial datasets. Analysis findings show that CNN model shows better results with 96.4 percent accuracy, which is higher than conventional machine learning techniques such as SVM and KNN. In addition, the system allows GUI-based operations to interact and scale easily, and enables the administrators to add new students, to train the model, and to generate reports without complications. Privacy is maintained because data is stored locally and it can be enhanced in the future to encrypted facial embeddings. There are also solutions to practical problems like proxy attendance, recognition when occluded (e.g. in face mask), and instructor feedback in real-time. In general, the solution offered is consistent with the objectives of digital transformation in smart campuses and offers a stable scalable framework that is flexible in various institutional configurations.