Apparent Age Estimation with CNN

Apparent age estimation from face image has become relevant to an increasing amount of applications, particularly since the rise of social platforms and social media. In this paper, we tackle the estimation of apparent age in still face images with deep convolutional neural networks (CNN). Our convolutional neural network use the GoogLeNet architecture, add batch normalization layer after each ReLU operation and remove all the dropout operations to accelerate the convergence of this very large-scale deep network. In addition, due to the limited number of apparent age annotated images, we train the deep models with several datasets in a cascaded way. Firstly, We pre-train a real age estimation model using IMDB-WIKI dataset, and then fine-tune the deep model with combined dataset with multiple real-age labeled databases. Finally, the apparent age data from the challenge are used to fine-tune the deep model parameters for apparent age estimation.