image generation via GANs
Python
Latest commit 9bfb030 Jun 29, 2016 @kvfrans links fixed
Permalink
Failed to load latest commit information.
data imagenet Jun 28, 2016
results adjusting z Jun 29, 2016
training new training with fixed z sampling Jun 29, 2016
trainingCIFAR imagenet Jun 28, 2016
.gitignore ignored Jun 26, 2016
README.md links fixed Jun 29, 2016
download.py imagenet Jun 28, 2016
main.py new training with fixed z sampling Jun 29, 2016
ops.py moved stuff around Jun 26, 2016
ops.pyc moved stuff around Jun 26, 2016
utils.py imagenet Jun 28, 2016
utils.pyc imagenet Jun 28, 2016

README.md

generative-adversial

Image generation via generative adversial networks on CIFAR-10, Imagenet, and Celeb faces written in Tensorflow.

This is code that goes along with my post about generative adversial networks.

How to use:

Download the datasets and place into the data folder.

Run main.py to start training, make sure to set train = True in the file.

Set train = False to visualize how adjusting the initial noise affects image generation.

Set cifar = True to train on CIFAR. This sets the network to train on 32x32 images instead of 64x64, and reads from the CIFAR binaries rather than from JPEGS.

Utils and network structure from DCGAN-Tensorflow.