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.
main.py to start training, make sure to set
train = True in the file.
train = False to visualize how adjusting the initial noise affects image generation.
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.