GENERATIVE ADVERSARIAL NETWORKS BASED IMAGE GENERATION PROCESSING DEVICE THAT ENABLES GENERATION OF IMAGES THROUGH MACHINE LEARNING WITH SUPPLEMENTARY DISCRIMINATOR AND OPERATING METHOD THEREOF

摘要

Disclosed are a generative adversarial neural network-based image generation processing apparatus and method that enable generation of images through machine learning to which auxiliary identifiers are added. The image generation processing apparatus and method according to the present invention primarily train an unlearned generator and an identifier based on a generative adversarial network (GANs) model, and learn in advance through a generator on which the primary learning is performed. After additional re-learning of the auxiliary identifier that has been previously learned, the generator and identifier on which the primary learning is performed and the auxiliary identifier on which the re-learning is performed are further trained based on the GANs model, eventually learning between the existing generator and the identifier It is possible to create a more sophisticated image in contrast to the technique of creating an image through this.