Application Research Of Face Age Prediction Based On Deep Convolutional Network

With the rapid emergence of smart applications,the need to automatically extract biological information from facial images is also growing.Then,applying the age prediction algorithm to various human-computer interaction systems related to age information can satisfy the application requirements in real life to a great extent.Based on deep learning,the representation of features can be automatically learned directly from big data.This paper mainly studies face age prediction based on deep convolutional neural networks.In order to predict the age of a face,it is first necessary to detect a face from the image in real time,and then perform age prediction based on the detected face image.The main research contents of this paper are as follows:A face detection framework based on concatenated convolutional neural network is designed and implemented for real-time face detection.The framework first preprocesses the face image to be detected by the image pyramid to extract face candidate windows of different scales.Then,the concatenated convolutional neural network takes the candidate face window as input,and eliminates the window without the face or the window containing the overlapping information by step-by-step fine filtering and non-maximum suppression.Finally,output the window containing the face and the five facial feature points of the face.The cascading convolutional neural network is designed to accurately capture the face window and key position,and to achieve real-time face detection.The experimental results show that the proposed face detection method is faster than the Viola-Jones and Haar face detection methods in terms of detection accuracy and speed.A face age prediction algorithm based on deep convolutional neural network is designed and implemented for the detected face images.The algorithm adopts a group convolution-based ResNeXt network architecture,which can improve the accuracy without increasing the parameter complexity,and also reduce the number of hyperparameters.For a given face detection image to be predicted,the designed face age prediction algorithm first extracts features by deep convolutional neural network,and outputs probability values belonging to 101(0 ~100 years old)ages,finally,the age value based on the weighted average output prediction.The experimental results show that the designed face age prediction algorithm can effectively predict the age.