Analysis And Research On Skin Quality Judgment Based On Deep Network Model

Deep learning(DL)technology has quietly entered all walks of life of human beings,and has been widely used in machine translation,object detection,automatic car driving and other fields by researchers.Deep convolutional neural network is its mainstream network structure,because of its powerful feature extraction ability,it has achieved remarkable research results in the field of image recognition and target detection.Nowadays,with the change of the concept of life health,people pay more and more attention to skin health,especially the quality of facial skin plays an extremely important role in personal image.Therefore,in view of the skin quality problems,the main research object of this paper is facial skin.It has certain research value and significance to use deep learning technology to accurately detect three kinds of skin problems in the skin,such as spots,papules and nevus,to help people make judgments on skin health.In this paper,deep learning technology is applied to the problem of skin quality judgment,and a skin quality judgment algorithm based on deep network model is proposed.On the basis of the deep network structure,an improved yolov3 deep convolution neural network is constructed to detect and recognize the common skin problems in the target area,and mark the specific location and category name of the abnormal area.The main research contents of this paper are as follows:1.Data preparation and preprocessing.The accuracy of data is very important for deep learning methods,so we should first understand and analyze the characteristics of existing data sets,and accurately classify and label the abnormal points on the skin;2.The selection of target area.The relative size of the object in the image affects the performance of the depth network model.In the basic backbone network,downsampling of features will result in less information contained in the feature graph of relatively small objects,poor network model learning effect,and ultimately reduce the performance of the model.Therefore,this paper proposes an improved YOLOv3 Tiny deep convolution neural network,which can quickly extract the face area,increase the relative size of spots,pimples and nevus in the input image,reduce the influence of background,and increase the final skin detection accuracy;3.A high performance loss function is constructed and a skin quality detection algorithm based on depth network model is proposed.The loss function reflects the difference between the predicted value and the actual value,which is one of the most important parts of the object detection algorithm.Based on the analysis of the position relationship between the prediction box and the target box,this paper proposes the Area IoU(AIoU)loss function,and applies it to the YOLOv3 object detection algorithm,and further proposes the skin quality detection algorithm,which completes the detection of abnormal skin points in the target area,and directly shows people the current skin health status.