Generative Adversarial Networks based Face Fractalization by using GAN

Abstract

Face Frontalization refers to generating the frontal view from a side faced view. There are a lot of crime scenes going on today in which a frontal face of the suscept is not perfectly visible. Even though many face recognition systems exist, it is still not possible to have a clear front view of the suspect. In order to find a clear face, the existing image should be rotated, this is where the face frontalization comes into action. This model will be able to rotate the available side face image in order to find a frontal face. To achieve this, the Generative Adversarial Networks (GAN) are used. The Generative Adversarial Network (GAN) consists of a discriminator and generator. The discriminator goes deep into the layers from the top layers to all the way leaving to the bottom most layers in order to get a deep understanding on the input image. The generator works in contrast to the discriminator and regain all the deep layers that are deconvoluted by the discriminator. Finally, both the generator and the discriminator will combinedly work to form a Generative Adversarial Network (GAN) and generate output based on the input image.