In this era of digital India, there is a change and transformation in every sector. In education, too, with more education becoming digitally based, real-time monitoring of student attention has become the key to improving learning competencies and interest. This paper proposes a real-time implementation of the deep learning object detection method YOLOv3 to create a system for the expectant real-time detection of engagement levels in learners. YOLOv3 is used as the tracking system for student engagement using facial expressions, including head position and direction of eye movement and gaze. Annotations were of high quality when training the models, due to using Roboflow datasets to preprocess the data and create a dataset. The features of engagement were captured in training the YOLOv3 model with precision, recall, mAP@50, and F1 score as measured performance metrics. For tracking students’ engagement, the YOLOv3 model was able to track student engagement with precision = 0.743, recall = 0.885, F1 score = 0.808 and mAP@50 = 0.857. The presented model was able to collect data at 25–30 FPS allowing a successful integration with the live class. Future improvements aim to enhance detection accuracy by integrating multimodal data fusion, temporal analysis, and more sophisticated deep learning approaches like transformers to extract more resilient feature sets. In conclusion, the system showcases an efficient and scalable solution to enhancing student engagement in online learning contexts based on real-time attention monitoring.