Abstract

The Image-to-image translation process belongs to a class of vision and graphics problem area, where the goal is to learn the mapping between an input and output image using a training set consisting of aligned image pairs. This research will focus on utilisation of image transitioning to make a Java Spring based Web-App which can transition a portrait into different human-oriented emotion (i.e., happy, sad, angry and frown), and simultaneously identify and transition its demographic region (i.e., America, Europe, Asia, South Asia) and same can be applied to the biological gender (i.e., Male, Female). The research study will use a proven deep learning methodology, which is Generative Adversarial Networks (GAN).

Keywords:General Adversarial Network (GAN); Deep learning; Pix-to-pix GAN; Cycle GAN; Webapp; Image-to-image transition