Draw the Desire: Bringing the sketches to life using Deep Learning

In this article, you will learn about conditional GAN (Generative Adversarial Network) and will be able to build one from scratch. After that you will be able to apply the cGAN model on a fashion products dataset for converting sketches of products to color images. If you would like to understand what are GANs [1] you can check out our previous tutorials on Latent Spaces. If we would like to generate one set of images while giving a different set of images as an input to a GAN model, this problem is called Image Translation. A classic GAN architecture doesn't take into account class labels therefore, we require a modified version of GAN.