Research Of Facial Expression Recognition And Face Recognition Based On Deep Learning Technology

With the development and maturity of deep learning technology,deep learning technology has been widely used in various research fields,making breakthroughs in many fields,especially in the field of computer vision.Image style transfer learning is a hot research direction in computer vision.Image style transfer can make the image transform the style while the image content remains unchanged.Style transfer learning trains the model mainly through two types of loss functions: content loss and style loss,so that the generated image style is transformed into the target image,while the content remains the same as the original image.Facial expression recognition technology belongs to the interdisciplinary research field of pattern recognition,emotion computing and other disciplines.The research of facial expression recognition is conducive to improving the efficiency of human-computer interaction,and human emotion can be obtained through computers.This technology can be applied to different scenes such as traffic,medical treatment and education.Face recognition technology has been successfully applied in people's real life,but the accuracy of face recognition is still affected by factors such as illumination,posture,expression and so on.Therefore,this paper applies style transfer technology to facial expression recognition and face recognition,and the main research contents are as follows:(1)Because a person's facial expression can be decomposed into expression components and neutral components,this paper proposes a facial expression recognition method based on style transfer.In this method,different generators are obtained by training Cycle-GAN.These different generators migrate different expressions to neutral,so that each generator corresponds to a different expression.In the test phase,the expression image is entered into the above trained generator.Facial expression recognition can be achieved in this way because only the generator corresponding to the input expression can migrate to neutral expression.The experimental results show that the new method not only performs well in the facial expression datasets obtained under laboratory conditions,but also has a very high recognition rate in the facial expression datasets obtained under natural conditions.(2)Inspired by the facial expression recognition method based on style transfer,a face recognition method based on style transfer is proposed.In face recognition tasks,multi-expression is a factor that affects the accuracy of face recognition.The method proposed in this paper uses style transfer technology to eliminate the influence of multiexpression,so as to improve the accuracy of face recognition.Specifically,Cycle-GAN is used as the style transfer model to train the generator that converts each expression image into neutral image in the data set.The converted facial expressions are applied to face recognition,and three different methods are used to extract facial features,and the experimental results are compared with those without the style transfer method,which proves that the proposed method has better performance on the data sets with large differences in facial expressions.