This repository is an implementation of model described in Pix2Pix (Isola et al., 2016).
The task is to translate picture to a different modality (e.g. day photo to night one) using conditional GAN. Model was trained on facades and day2night datasets. Those can be found here.
There are some good examples of trained models:
Also there were some discoveries and experiements, e.g.:
On the picture above dashed line represents loss on validation set. It's clearly seen that augmentations help to avoid overfitting.
More experiements can be found in my report (in Russian).