Disentangling Multiple Conditional Inputs in GANs

  • 2018-06-20 16:15:37
  • Gökhan Yildirim, Calvin Seward, Urs Bergmann
  • 5

Abstract

In this paper, we propose a method that disentangles the effects of multipleinput conditions in Generative Adversarial Networks (GANs). In particular, wedemonstrate our method in controlling color, texture, and shape of a generatedgarment image for computer-aided fashion design. To disentangle the effect ofinput attributes, we customize conditional GANs with consistency lossfunctions. In our experiments, we tune one input at a time and show that we canguide our network to generate novel and realistic images of clothing articles.In addition, we present a fashion design process that estimates the inputattributes of an existing garment and modifies them using our generator.

 

Quick Read (beta)

loading the full paper ...