Student Behavior Recognition in Classroom using Deep Transfer Learning with VGG-16

Tracking numerous students’ behavior by observing and questioning them is a difficult task. Therefore, several methods based on automatic facial expression recognition have been proposed to capture and make a summary of students’ behavior in the classroom. However, these methods cannot guarantee an effective classification due to the lack of huge datasets in this field. To improve students’ behavior identification from video sequences, we propose in this paper a new approach based on deep transfer learning. Our approach pre-trains the model on a facial expression dataset. Then, it transfers the model to classify students’ behavior. Experimental results confirm that our approach ensures a preferment students’ behavior classification.