Age Group Classification Based on Facial Features Using Deep Learning Method

Age group classification from a single face image using artificial intelligence applications such as virtual surveillance and marketing intelligence are a growing research area. Facial features extraction phase in a facial recognition classification model is one of the major phases as in this phase, facial elements like eyes and nose were extracted to distinguish a different person's facial features. Subsequently, this research aimed to design and build suitable age group classification model using Convolutional Neural Network (CNN) technique, utilizing the well-known, open-source UTKFace dataset. There were twelve customized CNN models built to find the best classification model for this research and every model could be differentiated by the number of convolutional layers and fully connected layers. In addition, there were two experiments conducted based on the observation of age group to further improve the performance for age group classification. These experiments were conducted by varying the number of convolutional layers, ranging from one to four convolutional layers, and three different batch sizes. Based on the analyzed results, the best classification model was the combination parameters of 64 batch size with two convolutional - three fully connected layers (2CNN-3FC) model, achieving a satisfactory accuracy of 71.40% for classification of 5 age groups (child, youth, adult, middle-age, and old). Future works for improvement and comparison purposes of the research can be made by experimenting with different deep learning method, and with other facial feature dataset.