Design and Development of a System for Monitoring Student Attention and Concentration during Learning Using CNN Model and Face Landmark Detection

—Mobile learning media has been wide and provides a tendency for lecturers to identify students' concentration levels in online classes. To bring the class into active learning, efforts are needed from lecturers and educational institutions to return students' concentration to the ongoing learning process. In this paper, a monitoring and alarm system is designed to increase student concentration and combines two elements of statistical analysis to validate CNN models that recognize face emotions in real time while learning. The research was carried out by recording face data using a camera, extracting digital features, and analyzing facial features. The results of the analysis are used as data input for the decision-making system regarding the level of concentration. The concentration level will be used to activate alarms and send them via chat so that students can focus on learning. The system is created by merging facial expression recognition (FER) and decision-making with a convolutional neural network. The system uses a face landmark via camera V2 and a Raspberry Pi 4 performed with the Haar-Cascade classifier, extracting facial features. Face detection via camera is performed using the Haar-Cascade classifier, which extracts facial features. The results of CNN model face detection with landmark features showed good results, with weighted average performance of precision, recall, and F1-score close to 0.99. According to the implementation results, the average number of facial expressions identified in drowsy and neutral states. The device can alert lecturers to how frequently drowsy detects students within a 10-minute interval.