demo program of Cycle-Shape-GAN using chainer

literature:
Translating and Segmenting Multimodal Medical Volumes with Cycle- and Shape-Consistency Generative Adversarial Network(https://arxiv.org/abs/1802.09655)

dependency

I confirmed operation only with..
1)python 3.5.3
2)chainer 3.3.0
3)cupy 2.3.0
4)Pillow 5.0.0

I cannot find MRI or CT images with annotation in internet, so I use cityScape dataset.
So, I modified several points to adapt that dataset...3d-conv to 2d-conv...etc