Cyclegan is a model to do 'sytle transfer'. It contains 2 generators and 2 discriminators
Genrator X : to generater a MAN's photo,based on a WOMAN's photo(condition)
Genrator Y : to generater a WOMAN's photo,based on a MAN's photo(condition)
discriminators X : to judge the MAN's photo, is it a fake one or real one?
discriminators y : to judge the WOMAN's photo, is it a fake one or real one?
Here's the training method:
1.real MAN-> gen x -> fake WOMAN -> dis y --> real or fake?
-> gen y -> fake MAN' --> loss with real MAN?
2.real WOMAN->gen y -> fake MAN -> dis x --> real or fake?
-> gen x -> fake WOMAN' --> loss with real WOMAN?