FERL-YOLO: Facial Expression Recognition Model of Learners

Abstract

Facial expression recognition of learners plays a crucial role in optimizing educational strategies. However, variability in facial expressions and environments limit the accuracy of existing models. To address this challenge, we propose FERL-YOLO, a model based on YOLOv11, integrating Haar Cascade for face detection, SENet for adaptive feature enhancement, and Optuna for hyperparameter optimization. Experiments on FER-2013 demonstrate that FERL-YOLO achieves competitive performance. Furthermore, to evaluate the reliability of emotion recognition, we conducted a reading experiment using our model to recognize learners’ emotions through facial expressions and learners’ self-reported feelings. Our findings provide valuable insights into the refinement of our model in future real educational settings.