Deep Age Estimation Using Sclera Images in Multiple Environment

Human age estimation from images using machine learning techniques is a challenging task. Due to physical aging process, color and texture of sclera, a protective outer layer present in human eye, get changed. In this work, we present an exploratory study to find the effectiveness of using sclera region of eye images for age estimation. It employs a modified form of deep neural network model VGG-16. The model is trained and tested by SBVPI dataset, in which the images are acquired with high-end cameras. The model is also tested using images acquired by a mobile camera fitted with a macro lens. The work gives the best mean-absolute-error of 0.06 and the encouraging results lead us to conclude that sclera images can be used as an effective modality for human age estimation. It is a pioneering work in the sense that the idea of using sclera for the purpose has not been explored before.