Student’s Engagement Detection Based on Computer Vision A Systematic Literature Review

Student engagement plays an essential role in shaping academic achievement, learning efficiency, and meaningful classroom interaction. As digital and hybrid learning environments continue to grow, conventional observation-based engagement assessment becomes insufficient, creating a strong need for automated and objective monitoring systems. This systematic literature review provides a comprehensive evaluation of computer vision–driven techniques for identifying and interpreting students’ engagement levels through visual cues.

The review examines approaches involving facial expression analysis, gaze-direction estimation, and body posture assessment, emphasizing how these behavioral indicators reflect emotional, cognitive, and attentional engagement. Advanced deep-learning architectures such as EfficientNetB0, MobileNetV2, and ResNet50 are highlighted for their capacity to extract high-level features, reduce computational overhead, and deliver robust performance in classroom and e-learning scenarios. These models demonstrate strong suitability for real-time engagement detection, especially in environments with variable lighting, spontaneous movement, and diverse student behaviors.

Additionally, the review discusses available datasets, data imbalance issues, preprocessing strategies, and commonly adopted metrics for model evaluation. Persistent challenges—such as occlusions, subtle facial variations, and generalization across different learners—are analyzed alongside the emerging trend of multimodal fusion, which combines facial cues, gaze, and posture for improved accuracy. Overall, this systematic study contributes toward developing intelligent, adaptive learning systems capable of supporting automated engagement monitoring and enhancing personalized educational experiences.

Keywords:

Student engagement, Computer vision, EfficientNetB0, MobileNetV2, ResNet50, Deep learning, Facial expression analysis, Eye-gaze estimation, Posture recognition, Multimodal assessment, Educational analytics, Systematic literature review.