Synthesis of Facial Image using Conditional Generative Adversarial Network

Abstract

Face sketch is done by sketch artist for a suspected or missing person from the description of an eyewitness. These methods have been widely used by forensic investigators. It is difficult for the sketch artist to draw perfectly from such verbal descriptions given by eyewitness of scenes and hard for the informer to confirm whether the sketch looks like the real person. In this work, we proposed a conditional generative adversarial network (cGAN) for synthesizing real human face taking a sketch as an input image. The focus of our model is to generate realistic images that preserve the identity the target person verified by face recognition algorithms. The proposed cGAN has been verified on a variety of facial sketches, which confirms the effectiveness and improved facial recognition score.