Great article!

There is a mistake in section 8 where you define the U-Net Conv2d-Block.

# second layer

x = Conv2D(filters = n_filters, kernel_size = (kernel_size, kernel_size),kernel_initializer = ‘he_normal’, padding = ‘same’)(input_tensor)

should be:

# second layer

x = Conv2D(filters = n_filters, kernel_size = (kernel_size, kernel_size), kernel_initializer = ‘he_normal’, padding = ‘same’)(x)