Hi @shrazi. this is the father of steemean
Jimbo went to sleep. :-) He asked me to answer your question about other Gaugan versions.?
Everybody can put on in a server two source code pieces of software.
SPADE - Semantic Image Synthesis with Spatially-Adaptive Normalization
https://github.com/NVlabs/SPADE for the first menu of Gaugan.
And together with the version 3 of PyTorch https://github.com/speedinghzl/Pytorch-Deeplab to convert pictures to segmentation maps- the second menu on Gaugan to import the photos.
The Gaugan uses version 2 PyTorch deeplabv2 under MIT license
that is much slower, :(
Pytorch - using photos on Gaugan- can find more colors of people cars, animals and so on, that you can later clone for the brushes to use in the paintings. :-)
See more in https://nvlabs.github.io/SPADE/