SVM-Classified FaceNet and Eigenface Models under Lighting and Occlusion Variations for Face Recognition

Emotional interactions between teachers and learners are essential in educational settings. This study aimed to help teachers recognize and respond to learners' emotions effectively by proposing a suitable face recognition system for use in learner emotion recognition systems. The research compared two models, FaceNet with Support Vector Machine (SVM) and Eigenface with SVM, using a single sample per learner. The results show that FaceNet and SVM models performed better, achieving almost perfect accuracy of 98%, while Eigenface only reached 84%. Under various lighting conditions, the FaceNet and SVM models maintained a high accuracy of 90%, whereas Eigenface reached only 9%. Even under partially covered face conditions, FaceNet and SVM recorded 67% accuracy, which was significantly higher than Eigenface and SVM's 11% accuracy. The adaptability of FaceNet under different lighting conditions and visual challenges makes it a suitable deep-learning model for emotion recognition in online learning.