Generative Models

This project is a collection of various flavors of Generative Adversarial Networks, popularly known as GAN. The implementation is done in TensorFlow.

About

Generative Adversarial Networks, or GAN, were first introduced by Goodfellow, et al. in a NIPS 2014 paper, and since then has sparked a lot of interest in the deep learning research community.

GANs are based on game theoretic scenario where two neural nets, called generator and discriminator, compete against each to reach a Nash equilibrium. The generator learns a mapping function from noise to the original images, while the discriminator learns to identify images that come from the real world, or from the generator.

What's in it?

1. Vanilla GAN
2. DCGAN

Dependencies

1. Python 2.7+ or Python 3.3+
2. TensorFlow 0.12.1
3. SciPy
4. Pillow