Skip to content
Tensorflow implement for Conditional convolutional adversarial networks.
Branch: master
Clone or download
zhangqianhui Merge pull request #6 from andrearama/master
Minor adjustment for for python 3
Latest commit 1b4d7ea Oct 10, 2018
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
images update Apr 5, 2017
.gitignore Initial commit Dec 12, 2016
LICENSE Initial commit Dec 12, 2016
README.md Update README.md Feb 24, 2018
download.py tf1.0 Apr 5, 2017
main.py update for python 3.6 Oct 4, 2018
make_gif.py code commit Dec 12, 2016
model_mnist.py Change initialization method for layer with sigmoid activation Oct 10, 2018
ops.py Change initialization method for layer with sigmoid activation Oct 10, 2018
utils.py update for python 3.6 Oct 4, 2018

README.md

Conditional-Gans

The test code for Conditional Generative Adversarial Nets using tensorflow.

INTRODUCTION

Tensorflow implements of Conditional Generative Adversarial Nets.The paper should be the first one to introduce Conditional GANS.But they did not provide source codes.My code has some differences comparing the paper:The Gans is based on Convolution network and the code refer to DCGAN.

Prerequisites

  • tensorflow >=1.0

  • python 2.7

  • opencv 2.4.8

  • scipy 0.13

Usage

Download mnist:

$ python download.py mnist

Train:

$ python main.py --op 0

Test:

$ python main.py --op 1

Visualization:

$ python main.py --op 2

GIF:

$ python make_gif.py

Result on mnist

Visualization:

the visualization of weights:

the visualization of activation:

Reference code

DCGAN

You can’t perform that action at this time.