Age Estimation Based on Integrated Convolutional Neural Network

In daily life, human faces due to factors such as lifestyle, work environment, and innate genes, showing different degrees of aging. Under the existing conditions, the face age estimation efficiency is low. This paper mainly studies face age classification under unfiltered faces. Unlike traditional face classification under standard images which are collected in laboratory environment, our age classification under unfiltered faces, has more realistic meaning and application value. This paper proposes an integrated algorithm for face age estimation based on convolutional neural network. Firstly, the image is preprocessed. The Preprocessing methods are data enhancement and histogram equalization. Then we design two different convolutional neural network structures. The feature is extracted by convolutional neural network. Finally, the age estimation result is obtained by the integrated algorithm. We have done experiments on the FGNET and CACD2000 datasets. Experiments show that the integrated algorithm, the accuracy is respectively 75.8% and 75.1%. The accuracy of NET1 is respectively 73.9% and 73.6%. The accuracy of NET2 is respectively 71.7% and 70.1%. The age estimation results infer that the integrated network has greater generalization ability and better robustness than the single convolutional neural network.