Research On Face Gender Recognition Algorithm Based On Deep Learning

In this paper,face recognition algorithm based on deep learning is studied.Face sex recognition means using computer to analyze face video or image,and extract effective information for identifying sex,and finally distinguish the gender information of the face object.The application of face gender recognition has been integrated into all aspects of life,such as visual monitoring,automatic identification,bank security,access security and other fields.First,this paper uses the images of men and women in the[IMDB-WIKI]data set as the data of the training model,and then gives the main techniques of classification based face recognition,including spatial transformation,linear interpolation and convolution neural networks in image processing.Then,the structure and implementation method of image feature extraction method using VGGNet,GoogleNet and ResNet using deep convolution neural network is given.Then,this paper presents a method of preprocessing the face image data.This method uses cascaded neural networks to filter candidate faces in the image,and uses convolution neural networks with different complexity to screen the candidate face sample regions at different stages,and the cascaded network structure will reduce a large number of samples into the final face region output.Then the face regions detected and recognized are sent to the classifier network to get the gender corresponding to face images.Because the pose of the face in the training sample is upright,so in order to improve the accuracy of the model recognition,this paper uses the facial feature points obtained by the cascade neural network to correct the face's face.On the basis of the structure of deep convolution neural network,this paper proposes an improved algorithm structure on the basis of the structure of deep convolution neural network,which can achieve reliable accuracy while achieving more reliable computing speed.Because of the small data set,too complex model structure and too many parameters,the model is difficult to converge.Therefore,a simple convolution neural network structure is designed in this paper for the face recognition task of this paper,and the performance of the algorithm is improved by a variety of deep neural network structures.It is found that the accuracy of the model can be significantly improved after using Drought and Batch norm technology.Finally,a face recognition system is designed.In view of the video sequence and the static picture,the face information in the image is detected in real time,and the detected face information is displayed in the picture to realize the dynamic monitoring and recognition of the face sex.