Automatic Age Estimation Research Based On Deep Learning

Face age estimation has become a hot topic in the field of machine vision.Many scholars at home and abroad have done a lot of research on it.But the change of age goes through extremely complex process.It is affected by the living environment,genetic factors and other factors.It is still a challenging task to predict the age by face image.The study of face age estimation can be divided into age feature extraction and age estimation.According to the situation of present research,this paper makes an in-depth study on the extraction of the age characteristics and the use of regression model to predict the age value.The main work of this paper is summarized as follows:1)A new feature learning method called PCANet feature learning model is proposed.PCANet learning model is a simple deep learning model.It studies the convolution kernel of volume layer through the PCA algorithm,and then extracts the age features of face images using this model.Deep learning architecture has been shown to have a very good learning ability.It is able to capture the characteristics of the face appearance changes and the characteristics with good description ability of the age.Therefore,the PCANet feature learning model can get better results in the age estimation of face images.2)This paper presents a method for expressing the characteristics of PCA network model with multilayer deep learning.The depth network is characterized by a layer by layer method,and the low level of the network can learn the edge features.The high level can learn abstract features.So the expression characteristics of multilayer PCA network model are the combination of low-level features and high-level features as the age characteristics of the final value.And the extraction of the age feature information is more abundant.3)An age estimation method of depth migration training model is proposed.In view of the fact that the training model of deep learning is likely to produce over fitting phenomenon under the condition of small data sets.It can effectively solve theover fitting problem.Firstly,the ImageNet image set is used to train the network model,and then the weights of the ImageNet image set are used as the initial weights of the network.Apparent face age data set is then used to fine tune the universal network model to make it have the ability to learn the age characteristics,so as to improve the effect of face age prediction.