Research On Classroom Head-up Rate Based On Deep Learning

With the development of science and technology,deep learning algorithms based on convolutional neural networks have achieved excellent results in computer vision,speech recognition,natural language processing,and other commercial fields.In particular,the use of deep learning technology to identify targets in pictures or live videos is also one of the current research hotspots in many industries.But at present,it is not common to apply its technology to the research of object recognition in classroom scenes,and the face and head posture of students in the teaching process of colleges and universities can reflect the students' listening status and the overall effect of the classroom.The current neural network technology has high recognition accuracy and fast speed.Therefore,it is necessary to apply deep learning technology based on convolutional neural network to college classroom scenes to provide reference for classroom evaluation and provide effective means for colleges and universities to improve the quality of classroom teaching.This article processes the videos collected from the teaching process of colleges and universities,and establishes a data set by manual annotation method,and annotates each face in each picture.Students with heads up are marked as heads up and students with heads down are marked as heads down.After comparing experiments with mainstream algorithms,the best Yolo-v3 algorithm is selected as the detection tool to train and test the data set,and the obtained network model is optimized.Adjust the size of the prior frame according to the clustering of your own data set,enhance the image data,and optimize the learning rate to increase the m AP of the original Yolo-v3 code from 84.51% to 87.32%.Aiming at the problem of face image recognition of small target students in the classroom,the receptive field module and attention mechanism are integrated into the network and the number of network feature extraction layers is increased to increase m AP from 87.32% to 90.07%.Finally,when predicting the image of the students' class status,the number of head-up students,the number of head-down students,and the head-up rate are added to the image display for teaching evaluation management in universities.Based on the improved Yolo-v3 head-up rate statistical system,the head-up rate statistics model can be used to calculate the head-up rate of students in courses,classes,classes,teachers,and colleges.In order to display the statistical results of the head-up rate more conveniently and intuitively,the historical and current statistical information is saved and displayed through the visual system module,which provides a reference for the evaluation of the classroom head-up rate in colleges and universities,and also opens up a new way for colleges and universities to evaluate students and teachers.Methods.