1D/2D keras U-net model.
unet = make_unet(input_shape=(64,64, 3),
nout=1, # number of channels in output
scales=5,
nconvs_by_scale=2,
base_filters=8,
kernel_size=3,
activation='relu',
first_activation='tanh',
last_activation='linear',
interpolator='nearest',
last_interpolator=None,
norm=False,
dropout=False,
norm_at_start=False,
nconvs_bottom=None,
use_skip_connections=True,
return_encoders=False,
verbose=False)
Y = unet.predict(X)