Aiming at the problems that the mask detection model lacks the detection of non-standard wearing classification,and it is difficult to be compatible with high precision and speed,a one-stage real-time detection algorithm for standard wearing of masks was proposed.The lightweight extraction network DM-CSP was introduced,and multi scale attention was added to enhance the extraction ability.Aiming at the problem that the deep and shallow feature information is not aligned in the fusion stage,FAS(feature alignment and screening)was designed,and CTM(feature enhancement module)was proposed to correlate the feature map context information.The decoupling channels were constructed for image recognition to further improve the reco-gnition accuracy and convergence speed of the algorithm.Experimental results show that the detection accuracy of the improved algorithm is 93.2％,which is 4.8％higher than that of the mainstream algorithm YOLOv4-Tiny.It has better performance in detection speed and model capacity.
维普
万方数据