So I created the two .tfrecords files and ran training with a database of about 800 A images and about 2000 B images.
Tensorboard shows training previews but it doesn't look like it's training X->Y.
It displays first X_input then X_generated then X_reconstruction, and below same for some other Y image. In both cases input and generated are more or less the same.
It looks as if it was learning X->X and Y-Y separately.
Where are the X->Y attempts? Should it look like that in initial stages of training?
Hi,
So I created the two .tfrecords files and ran training with a database of about 800 A images and about 2000 B images.
Tensorboard shows training previews but it doesn't look like it's training X->Y. It displays first X_input then X_generated then X_reconstruction, and below same for some other Y image. In both cases input and generated are more or less the same. It looks as if it was learning X->X and Y-Y separately.
Where are the X->Y attempts? Should it look like that in initial stages of training?