Gan_Architecture

create an architecture for Generative Adversarial Networks. implementations

Table of Contents

• Installation
• Implementations
  • Spectral Normalization
  • GAN with Info

Installation

$ git clone https://github.com/VitoRazor/Gan_Architecture.git
$ cd Gan_Architecture-master/
$ pip install keras

Implementations

GAN-SN

Implementation of Generative Adversarial Network with Spectral Normalization for Wasserstein-divergence

Code

Reference Paper:

Spectral normalization for generative adversarial networks:https://arxiv.org/abs/1802.05957

Wasserstein GAN: https://arxiv.org/abs/1701.07875

Result: Train fro cartoon characters 64x64

Train fro aerial image 64x64[iteration=150000] and 256x256[iteration=34800]

GAN-info

Implementation of Generative Adversarial Network with InfoGAN and ACGAN, simultaneously using Spectral Normalization for Wasserstein-divergence.

Code

Reference Paper:

Auxiliary Classifier Generative Adversarial Network: https://arxiv.org/abs/1610.09585

Interpretable Representation Learning by Information Maximizing Generative Adversarial Nets: https://arxiv.org/abs/1606.03657

Result: from iteration 10 to iteration 15000