| The quality of school education affects the growth of people,to a certain extent the progress of society.It’s got much attention by society.Classroom teaching is the main form of school education.Due to the lack of objective and effective evaluation methods,there are currently problems such as difficulty in evaluating classroom teaching quality,lack of objective evaluation evidences for teaching process management and quality monitoring.The development of artificial intelligence technology has provided new ideas for classroom teaching evaluation.But there are still a series of problems in the classroom evaluation scheme on the market such as high system cost,low evaluation accuracy and incomplete evaluation.In view of the problems above,the thesis proposes a solution for classroom evaluation system based on deep learning.The solution reseaches the evaluation methods of students’ concentration,classroom activity and teaching links,and builds an informationize evaluation system for students’ learning process and classroom teaching quality.The thesis uses the system architecture of edge computing,builds a hardware platform with an FPGA deep learning acceleration server to manage multiple AI plus embedded visual edge computing devices with PTZ,and completes the design,debugging and testing of classroom evaluation and students’ behavior statistics functions based on the platform.This thesis uses the edge computing hardware architecture to solve the problem of high system cost,uses the deep learning technology to solve the problem of low classroom evaluation accuracy,and solve the problem of incomplete classroom evaluation by extracting multiple indicators such as students’ attendance,classroom concentration,classroom activity and richness of teaching links to objectively evaluate the classroom.The innovations of the thesis are as follows:(1)For a single fixed camera cannot achieve full classroom coverage,and the inconvenience of multi-camera wiring,the thesis uses intelligent scanning to achieve efficient classroom area monitoring.(2)For the problem of limited system bandwidth,the thesis uses the edge computing architecture to allocate high data-dependent operations such as image segmentation to the edge side,face recognition and other calculations with less data-dependent to the server,which can reduce bandwidth overhead and increase system capacity.(3)For the difficulties in teaching process management and quantitative evaluation of teaching quality,the thesis uses deep learning technology to make semantic evaluation of classroom learning,intelligently extract classroom related information,which achieve intelligent evaluation and information management of classroom teaching quality.After testing,the classroom evaluation system designed in the thesis is stable,with every functions running well.According to the test results,the system can basically meet the application requirements of classroom teaching evaluation. |
