Deep Convolutional GANs or DCGANs are specific types of GAN in which the generator and/or discriminator are build/trained using convolutional layers. The DCGAN architecture was first presented and tested by Alec Radford, Luke Metz and Soumith Chintala in this paper.
In this repository, I'm going to include my work on training a DSGAN on the Street View House Numbers (SVHN) dataset . The images in this dataset are collected from Google street view. The goal is to create a DCGAN that can hopefully generate new, realistic-looking images of house numbers.