Detection of Student Sleepiness Using U-Net and Conv-LSTM Methods to Improve Learning Quality

In the context of monitoring student engagement during learning sessions, the ability to accurately detect sleepiness can be a valuable asset. Early detection of sleepiness is critical to evaluate the need for intervention to ensure effective learning outcomes. Prolonged student sleepiness can negatively impact learning effectiveness and may result in academic challenges. This research used a semantic segmentation network to monitor and detect the level of sleepiness in students during learning sessions by modifying the U-Net architecture, which is enhanced with Convolutional Long Short-Term Memory (Conv-LSTM) blocks. Conv-LSTM blocks enable efficient feature sharing between network layers and improve accuracy in detecting sleepiness. In this study, experiments were carried out and the proposed method for monitoring students in the classroom was evaluated. The results of this research showed that the percentage of students who made movements detected in the video and extracted in image detection was 92% accurate and not sleepy.