rev = transforms_ft.decode(out)
show_images((rev[0][0], rev[1]))