"ValueError: None values not supported" when calling SubpixelConv2D function

Question

I am trying to utilize the SubpixelConv2D function

I am training a GAN and wanted to up sample using subpixel instead of interpolation or convolution transpose due to the artifacts they leave behind.

I am using Tensorflow/1.4.0 and Keras/2.2.4

When I try and call the function I receive the following error:

"ValueError: None values not supported."

I call the function using :

import tensorflow as tf
from tensorflow import keras
import Utils

def up_sampling_block(model):

    #model = keras.layers.Conv2DTranspose(filters = filters, kernel_size = kernal_size, strides = strides, padding = "same")(model)
    model = Utils.SubpixelConv2D(model)(model)
    #model = keras.layers.UpSampling2D(size = 2)(model)
    #model = keras.layers.Conv2D(filters = filters, kernel_size = kernal_size, strides = strides, padding = "same")(model)
    #model = keras.layers.LeakyReLU(alpha = 0.2)(model)

    return model

and the function is as follows:

# Subpixel Conv will upsample from (h, w, c) to (h/r, w/r, c/r^2)
def SubpixelConv2D(input_shape, scale=4):
    def subpixel_shape(input_shape, scale):
        dims = [input_shape[0], input_shape[1] * scale, input_shape[2] * scale, int(input_shape[3] / (scale ** 2))]             
        output_shape = tuple(dims)
        return output_shape

    def subpixel(x):
        return tf.depth_to_space(x, scale)

    return keras.layers.Lambda(subpixel, subpixel_shape)

The size of the input tensor is (?,48,48,64), and I believe the "?" for the batch size is causing the error, but I can't seem to fix the problem.

Answer 1

The second function of the Lambda layer has to be a function of the input shape only: subpixel_shape(input_shape) but your takes a second argument called scale that default to undefined when only input_shape is passed. Try passing lambda input_shape: subpixel_shape(input_shape, scale) to keras.layers.Lambda function instead. Then scale will default to 4 as the outer function dictates. Or remove the scale from the subpixel_shape function arguments:

def outer(a=0):
    def inner():
        print(a)

    return inner

print(outer()()) # prints 0