polyGAN

polyGAN is a Generative Adversarial Network(GAN) based on the DC-GAN architecture. It is implemented using Keras in Python. polyGAN is capable of training on and generating color images as well as grayscale images of any given resolution.

It is implemented in Python using the Keras library and tested on MNIST, CIFAR10 and Chars74K image datasets.

Architecture

GAN

Deep-Convolutional GAN (DC-GAN)

Results

Using the MNIST dataset

Using the CIFAR10 dataset

Using alphabet images from the Chars74K dataset

References

Radford, Alec, Luke Metz, and Soumith Chintala. "Unsupervised representation learning with deep convolutional generative adversarial networks." arXiv preprint arXiv:1511.06434 (2015).

Dependencies

Python 3.5
Numpy
Tensorflow
Keras
OpenCV