now, in production....
Implementation of probabilistic U-Net model using tensorflow. If any bug, please send me e-mail.
I confirmed operation only with..
1)python 3.6.3
2)tensorflow 1.7.0
3)numpy 1.14.2
4)Pillow 4.3.0
metrics and etc....
after 100 epochs, I have predicted CityScape test datasets.
1st row: original image(cityScape datasets test images).
2nd row: labels.
3rd row: prediction.
and, predict 1 test image changing latent z.
