Design And Implementation Of Class State Feedback System Based On Face Pose Recognition

Face recognition technology is one of the hot research directions in computer vision,and has been applied in many fields.At the same time,"rise rate" has become one of the important factors to judge students’ learning concentration and test the effect of classroom teaching.Therefore,a class state feedback system based on face pose recognition has designed and developed in this paper.Firstly,the research status and significance of face recognition technology and vision based classroom teaching evaluation method at home and abroad has been instructed in the paper.Secondly,the function and implementation effect of convolution neural network used in this algorithm has been expounded,and the face recognition principle and loss function of yolov5 s has been summarized in the paper.Then the algorithm of face location and recognition is designed and implemented.Through the production of data sets,image preprocessing,the analysis of the results of initial training,as well as the improvement of the algorithm and the analysis and comparison of the improved results,the combination of various aspects to achieve face posture positioning recognition.The recognition accuracy of the initial training results of the yolov5 s training model is 0.985,which can not meet the application requirements of this subject.The model is improved in the process of algorithm implementation.Under the condition of ensuring its speed,the parameters in the yolov5 s training model are adjusted as a whole,and the depth and width of the model are increased to form a new model system.The data set is trained again.After the comparison of the recognition results before and after recognition,it is found that,the improved model maintains the speed advantage of the original model,and the recognition accuracy reaches 0.996.In order to detect the recognition effect further,the recognition test on the image in dark environment has been performed in the algorithm,which is found that the recognition result can reach 0.877,and it can meet the application requirements.On this basis,the value of identify parameters such as Conf and IOU parameters has been adjusted in the algorithm,so that the identification results are more clear.The experimental results showed that the accuracy of the improved system is improved by more than 0.5,and the average speed of face detection is less than 1.3seconds,it can quickly identify the students’ listening state in class,which meets the requirements of real-time monitoring in class teaching,and the confidence level of the occluding face recognition can reach 0.6 or above,which meets the needs of face recognition under special circumstances,the study provides the basis for the classroom student study condition monitoring.Finally,the summary and the forecast are expounded.The learning content,research,design and development of the Algorithm are summarized,and the future related work is pointed out in the paper.