Age Estimation from Facial Images using Transfer Learning and K-fold Cross-Validation

Automatic Age estimation has gained more and more interest in recent years due to its potential in many applications. Most techniques uses hand-crafted features to predict aging patterns, but not accurate enough to be employed effectively. Recent advances in deeply learned features extracted by Convolutional Neural Network (CNN) allows to design more accurate facial analysis. The aim of this paper is to explore the performance of different age estimation techniques that uses Deep Learning methods and to propose a variation of Transfer learning which uses K-fold cross validation on top of transfer learning. The experiment was carried out with UTKFace dataset using VGG16, ResNet50 and SENet50 models. The result demonstrates that our method is Superior to the existing methods in terms of performance.