Research And Practice Of Classroom Concentration Based On Face Detection Technology

The classroom is the display window of wisdom education transformation,and the effect of curriculum wisdom transformation also needs to be reflected in every class.The Twentieth National Congress of the Party proposed the strategic goal of promoting Chinese-style modernization and creating a modern university lecture system with Chinese characteristics.Currently,in the classroom scenario with many students in the class,teachers cannot monitor each student’s listening situation in real time,which in turn is likely to cause problems such as inefficient classroom management,lack of personalization in teaching process management and imperfect student evaluation mechanism.To address the above problems,this paper conducts a study on student face detection and face expression direction in the context of student-intensive classroom from the perspective of realizing classroom wisdom,in order to improve the effectiveness of teachers’ real-time monitoring and evaluation of the classroom.The main research of this paper is as follows.(1)Face detection algorithm improvement: For the classroom scenario where faces are dense and obscure each other,and more small targets are easy to cause missed detection,the YOLOv5 s algorithm in the target detection domain is improved by first replacing the backbone network of the original YOLOv5 s model with EfficientNet-Lite to reduce the size and computation of the model;then replacing the NMS with DIoUNMS non-maximal value suppression algorithm to improve the accuracy when faces are crowded,as well as introducing the CBAM attention mechanism to make it focus more on key information to improve the detection accuracy.And the effectiveness of the improved YOLOv5 s face detection algorithm is demonstrated experimentally.(2)Expression recognition algorithm improvement: Based on the VGG-19 network structure,we propose a dual-channel BC-Block module by fusing the residual structure and the bottleneck structure,and add a batch normalization layer after the convolutional layer in the original model as well as replace the original fully connected layer with a global average pooling GAP layer.Experimental validation in crosssectional and longitudinal comparison experiments as well as real classroom scenarios demonstrate the effectiveness of the improved model in expression recognition in this paper,which can identify common expressions of students in classroom environments and thus determine their real learning emotions.(3)System design and practice: According to the improved model above,the classroom assessment axis of "classroom head-up rate-student expression-classroom concentration" is established.Through the analysis of system requirements,this paper designs and practices a classroom concentration system based on face detection technology and applies it in the classroom.The system can effectively assist teachers to accurately grasp the classroom concentration and improve the effectiveness of classroom teaching.