This is an example implementation of DCGAN (https://arxiv.org/abs/1511.06434) using trainer.
This code uses Cifar-10 dataset by default.
You can use your own dataset by specifying
--dataset argument to the directory consisting of image files for training.
The model assumes the resolution of an input image is 32x32.
If you want to use another image resolution, you need to change the network architecture in net.py.
Below is an example learning result using cifar-10 dataset after 200 epoch.