Improvements made for training Compositional Pattern Producing Network as a Generative Model, using Residual Generative Adversarial Networks and Variational Autoencoder techniques to produce high resolution images.
Run python train.py from the command line to train from scratch and experiment with different settings.
sampler.py can be used inside IPython to interactively see results from the models being trained.
See my blog post at blog.otoro.net for more details.
I tested the implementation on TensorFlow 0.80.
Used images2gif.py written by Almar Klein, Ant1, Marius van Voorden.
BSD - images2gif.py
MIT - everything else