This repository is about Machine Learning Challenge 2017 onsite round task, but contains the model which I implemented and trained after the contest.
The source code is originally from mlc2017-onsite and the further information can be found in mlchallenge2017(ko) and google blog korea
The contest is to make a GAN(Generative Adversarial Network) and compete with the other participants.
My teamname was etude and the result can be found here.
My model is based on DCGAN(Deep Convolutional Generative Adversarial Network) with some tweaks.
After training this model (about 90k steps), some (fake) images from the generator are as follows.
While the original (training images or real images) images are as follows.
