Based on improved yolo_v3 for college students’ classroom behavior recognition

Abstract

The main purpose of the research on students’ classroom behavior recognition is to further systematically count all kinds of behavior data of students in class, and to provide a reliable technical support for education and teaching evaluation. Nowadays, the mainstream of target detection and recognition for multiple students in the classroom is to use various target detection and recognition technologies based on deep learning methods. These technologies optimize the model through self-learning of the data set through deep convolutional neural networks, thereby further Improve recognition efficiency. With the development of deep learning technology, the recognition efficiency has been greatly improved from the initial two-step detection to the current single-step detection algorithm. In the complex environment of the classroom, it is difficult to recognize students’ classroom behavior, which is effectively the problem of insufficient small target recognition accuracy. The original yolo-v3 network model is improved to make it suitable for students’ classrooms, which can solve this problem very well. According to the data fed back from the experimental results, the improved model has greatly improved the recognition efficiency.