Student Classroom behavior Recognition based on OpenPose and Deep Learning

With the education into the intelligent era, intelligent student classroom behavior recognition is becoming more and more important. However, it is difficult to identify students' classroom behavior intelligently because of the complexity and variability of students' classroom behavior. In order to improve the accuracy of intelligent student behavior recognition, seven typical classroom behavior images of 300 students are collected and the data are preprocessed. Then, the classical deep network model VGG16, which has been trained on ImageNet dataset, is transferred to student classroom behavior recognition task. Finally, through the experimental comparison with other deep learning models, the paper verifies that VGG16 network model has high recognition accuracy for students' classroom behavior. The above research shows that the identification of students' classroom behavior based on deep learning can provide timely and accurate feedback on students' classroom learning, which is conducive to teachers' improvement of teaching methods, optimization of classroom teaching and management, thus improving the efficiency of teaching and learning and facilitating teaching reform.