Face Detection in Digital Image Using Convolutional Neural Network Method for Web-Based Class Attendance System

Abstract

The main problem in this research is the need for a more efficient and accurate class attendance system. Manual attendance is often inefficient and has the potential for error. At Universitas Negeri Padang, especially at the Pesisir Selatan Campus for the informatics study program class of 2023, attendance is carried out through portals, e-learning, and in writing on paper during class, which can cause data mismatches, human error, and the possibility of data manipulation. This research aims to develop a face detection-based automatic attendance system using the Convolutional Neural Network method. This system is proposed to replace the less efficient and error-prone manual attendance method. This research uses the CNN method trained with a face dataset that includes variations in lighting and pose. From 800 face images of 32 students, the results show that the system has a face detection accuracy rate of 90% for face detection using glasses, and 100% accuracy for face detection without glasses. The system enables the attendance process to be faster and more accurate, increasing efficiency. In conclusion, the CNN method is proven to be effective in face detection on digital images and can be well implemented in a class attendance system, thus improving the efficiency and accuracy of the attendance process.