Cloud based Employee Timesheet Model Using Facial Recognition; in case of Oromia Occupational Competence Assurance Agency

Abstract

This paper introduces a cloud-based employee timesheet model that utilizes facial recognition methods, specifically targeting the Oromia Occupation Competence Assurance Agency. The limitations of traditional paper-based timesheets are highlighted, and a facial recognition-based model is proposed as a solution. The motivation behind this work stems from the inefficiencies, time consumption, and inaccuracies of the traditional timesheet process, exacerbated by the need for touchless solutions during the COVID-19 pandemic. Leveraging facial recognition, a biometric technology that identifies individuals based on unique facial features, presents a promising alternative. The system incorporates the Haar Cascade Classifier for face alignment and detection, followed by a Convolutional Neural Network (CNN) for extracting high-level features and matching faces in a cloud-based database. By combining computer vision algorithms, facial recognition technology, and cloud-based storage, the proposed solution aims to streamline the timesheet process and enhance efficiency. The study collected datasets from publicly available face databases from Flicker database and supplemented them with suitable Ethiopian face datasets. A total of 2945 datasets were utilized, with 441 known faces and 2504 unknown faces. The data was split into training, validation, and testing sets, employing holdout classification to avoid overfitting while it helps in assessing how well the model generalizes to unseen data. The performance of algorithms such as CNN, ANN, and SVM was compared, with CNN achieving the highest training accuracy (97.83%) and validation accuracy (88.30%), along with the lowest training and validation losses. Consequently, CNN was identified as the most suitable model for the timesheet system.