基于深度学习的到课率统计系统设计与实现

Design and Implementation of a Class Attendance Statistics System Based on Deep Learning

到课率作为宏观教学管理数据,对高校教学管理具有重要作用。虽然近年来出现了一些课率统计的数字化方法,解决了传统到课率统计费时、费力、滞后等问题,但由于成本高、使用不方便、准确率不高等原因,导致其无法推广。随着技术的发展,深度学习在多目标检测中的准确率越来越高,有助于解决此类问题。为此,文章利用深度学习技术,设计了一种基于教室摄像头RTSP视频流的到课学生头部识别的模型1MB-Plus,并将其应用于某高校的一百余间教室的到课率统计中,取得了97.3%的准确率。研究表明,该模型有助于解决到课率统计存在的问题。文章通过研究,旨在以最小的成本为高校教务管理部门提供较为准确的宏观到课率数据,辅助学校的教学管理工作。

As a macro teaching management data,class attendance rate plays an important role in the teaching management of colleges and universities.Although some digital methods of class attendance rate statistics methods have emerged in recent years to solve the problems of time,effort and lag in traditional class attendance statistics,they cannot be popularized due to high cost,inconvenient use and low accuracy.With the development of technology,the accuracy of deep learning in multi-target detection is increasingly higher,helping to solve such problems.Therefore,this paper used deep learning technology to design a 1MB-Plus model based on classroom camera RTSP video stream to recognize arriving students’heads,and applied it to the class attendance statistics in more than 100 classrooms of a university,and obtained the accuracy of 97.3%.The experiment also showed that the model was helpful to solve the problems of class attendance statistics.Through research,this paper was aimed to provide more accurate macroscopic attendance data for the educational administration departments of colleges and universities at the lowest cost,so as to assist the school teaching management.