Did you run into any such errors?

ValueError: A target array with shape (8, 512, 512, 3) was passed for an output of shape (None, 512, 512, 1) while using as loss binary_crossentropy. This loss expects targets to have the same shape as the output.

In your up_block the convolution between upsampling and concatenation is missing. This is why there non-power-of-two shapes in your log.

It should look like this:

def up_block(x, skip, filters, kernel_size=(3, 3), padding="same", strides=1):
    us = keras.layers.UpSampling2D((2, 2))(x)
    cv = keras.layers.Conv2D(filters, kernel_size, padding=padding, strides=strides, activation="relu")(us)
    concat = keras.layers.Concatenate()([cv, skip])
    # ...

What if i wanna use transfer learning for this code? like VGG16 or ResNet50?