<![CDATA[This study aims to develop and evaluate an efficient and accurate face embedding-based attendance system to address the limitations of the fingerprint-based attendance system still in use at Ichsan Gorontalo University. The system was developed using the FaceNet model for 512-dimensional face embedding extraction, with facial similarity comparison performed using Cosine Distance. The system development followed the Waterfall methodology, encompassing analysis, design, implementation, and testing phases. Testing was conducted through three approaches: White Box Testing to evaluate programming logic, Black Box Testing for functional validation, and User Acceptance Testing (UAT) to measure user satisfaction. Accuracy testing was performed under three different conditions involving 10 volunteers (7 registered, 3 unregistered): neutral facial expression (at 1 meter distance), smiling expression (at 1 meter distance), and 5-meter distance. The results demonstrate that the system exhibits low logical complexity with a Cyclomatic Complexity (CC) value of 7, all functional components operate without significant errors, and it achieved a user satisfaction rate of 84.53% (Grade B). Accuracy testing yielded 90% accuracy under both neutral and smiling expression conditions, but decreased to 60% at 5-meter distance. The system achieved an average response time of 1.2 seconds with memory usage below 2 GB. This study concludes that the face embedding-based attendance system is effective and efficient for use under normal facial expression conditions and close-range scenarios, and is recommended for implementation as a more accurate and hygienic modern attendance solution.]]>
Copyrights © 2026
