基于深度神经网络的口罩佩戴检测

Mask Wearing Detection Based on Deep Neural Network

深度神经网络在物体识别和分类中应用广泛,将其用于口罩佩戴检测,有利于提高新冠疫情防控管理工作效率。首先,收集佩戴口罩图片,将样本图片数据集扩充到12 000张。然后用Pytorch搭建ResNet-34深度神经网络,经适当预处理,调整学习率大小和批数据量大小,网络在验证集上准确率为98.41%,在测试集上准确率为97.25%。该网络对单张图片的检测用时为0.103秒,拥有较高的检测准确率和效率,能够满足公共场所对口罩佩戴检测的应用需求。

Deep neural network is widely used in object recognition and classification.A mask wearing detection method based on deep neural network is benefit to improve the efficiency of COVI-19’s prevention and control management.First,images of masks were collected and the sample data set was expanded to 12,000 images.Then,Pytorch was used to build the ResNet-34 deep neural network,and the network detection accuracy reached 98.41%on validation set and 97.25%on test set by optimizing the network learning rate,batch size and image preprocessing method.The detection time of this network is0.103 seconds for a single picture,which has a high detection accuracy and efficiency,and can meet the application demand of wearing masks in public places.