Online and hybrid learning require support systems that can assist educators in assessing student engagement more objectively. Student engagement cannot be adequately understood solely through attendance or click-based activity, as engagement encompasses behavioral, cognitive, emotional, and social dimensions. This study develops a real-time student engagement detection application based on facial expressions by integrating MediaPipe Face Detection with the Mini-Xception model. The FER-2013 dataset was used to train a seven-class facial expression classification model, comprising angry, disgust, fear, happy, neutral, sad, and surprise. The resulting facial expression classifications were subsequently mapped into two engagement categories: Engaged and Disengaged. The research methodology comprised image preprocessing, data augmentation, Mini-Xception model design, model training, classification evaluation, engagement mapping evaluation, and computational resource utilization monitoring.