Facial Expression Detection System for Students in Classroom Learning Process Using YoloV7

The utilization of technology in education is not only about using hardware or software, but also how technology can facilitate effective learning experiences. However, in the learning process there is a problem for teachers to know the level of student attention in the classroom to the material presented, so that the teacher does not know accurately the concentration of students during the learning process until it has an impact on the teacher's learning methods that are not in accordance with the characteristics of students. The purpose of this research is to detect students' facial expressions in the classroom learning process using yolov7. The implementation of several architectural models on CNN consists of several proposed methods, namely data collection, data augmentation, data annotation, split dataset, training, and model evaluation. System testing is done by measuring accuracy and comparing with other methods, namely CNN, CNN MobileNet, CNN EfficientNet-B0 and YoloV7. The test results show the average accuracy of CNN 80%, CNN MobileNet 93%, CNN EfficientNet-B0 31% and YoloV7 96%. Based on these results, it can be concluded that the YoloV7 method can detect student concentration effectively and efficiently compared to CNN, CNN MobileNet, and CNN EfficientNet-B0.