Research On Face Detection Algorithm Based On Deep Learning

Face detection has always been a hot topic in computer vision research.In recent years,with the development of the Internet of things,facial features,as a kind of identity information,are widely used in all kinds of automation and internet systems,such as electronic identity authentication,target monitoring,security alerts and other kinds of Internet systems,which are an important part of the future intelligent social ecology.The purpose of face detection is to find faces in pictures or videos,and mark the size and location of faces with tags.In actual detection,complex scenes and different scales have brought many difficulties to face detection,which leads to the traditional detection methods can not be used in the actual scene.The application of deep learning has brought a revolution to face detection technology.Using the powerful fitting ability of deep neural network,the accuracy of face detection algorithm has been greatly improved.However,these high precision algorithms can not meet the real-time detection requirements at the speed of detection,and can hardly be applied to the actual scene.In this paper,we hope to build a new full convolution neural network in the face detection algorithm based on the depth learning theory and on the existing high precision deep neural network model.It can further improve the detection accuracy while ensuring the detection speed.In this paper,the traditional face detection algorithm and the face detection algorithm based on depth learning are discussed.The two representative algorithms of Adaboost and FasterRCNN are introduced.In view of the shortcomings of these two algorithms,a real-time face detector is trained on the basis of the universal real-time target detection algorithm YOLOv2.The face detector is fast and the detection rate is much higher than that of the traditional algorithm,but its accuracy is not much better than that of FasterRCNN.Therefore,according to the requirements of the actual application,the existing convolution neural network model is improved,the faceyolo face detection algorithm based on the residual network is proposed,and based on the loss function and image enhancement,the FDDB data and the data are compared and tested.It is proved that faceyolo is in the original high-speed detection.It improves the accuracy of detection and improves the detection effect of small face and multiple faces.Finally,based on the faceyolo algorithm,this paper constructs a face detection system which can carry on the image face detection and call the real-time face detection of the camera.It can test the detection effect of the algorithm,and also can be put into practical application.