Train Compositional Pattern Producing Network as a Generative Model, using Generative Adversarial Networks and Variational Autoencoder techniques to produce high resolution images.
Truck class of CIFAR-10
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 on training on the MNIST set.
This version is an experimental hacked version of the MNIST model to train on CIFAR-10. Results are not really good yet, but I decided to just put the code up in case anyone wants to play with it and make it work.
I wrote another blog post about some of the current generative results on CIFAR-10, and what I think can improve this model going forward.
Tested this implementation on TensorFlow 0.60.
Used images2gif.py written by Almar Klein, Ant1, Marius van Voorden.
BSD - images2gif.py
MIT - everything else