Image translation with multidirectional autoencoders that share a latent space.
Comprehensive documentation will follow...
The data and the corresponding label image need to have the same size.
/dataset/
----/trainA <- data_x.png data part of the image translation
----/trainB <- label_x.png label part to the corresponding data_x.png in trainA
----/testA <- data for testing, otherwise same as trainA
----/testB <- labels for testing, otherwise same as trainB