GENERATIVE ADVERSARIAL NETWORKS AND APPICATION FOR BUILDING CHARACTERS IN VIRTUAL REALITY

Abstract

The article explored Generative Adversarial Networks (GAN) neural networks and application of automated material for human characters in virtual reality. A real data set of materials created by 3D designers was used to train two opposing elements in a neural network, which are data generating and data differentiating. Experimental results show that the GAN network allows automatic material generation for 3D models. GAN generated materials meet the technical requirements of meshing and imagery for use in 3D character modeling. This is the direction of research and application of artificial intelligence with great potentials in the production of multimedia data in general and 3D models in particular.