Reshape incorrect for Style Transfer

[TimoFriedri]

Hi,

I used your code coming from this article:
towardsdatascience

In your Notebook "Neural Style Transfer.ipynb" you have the following lines:

a_S = tf.reshape(a_S, [n_C, n_H * n_W])
 a_G = tf.reshape(a_G, [n_C, n_H * n_W])

This is the same mistake I did.
I think it has to be like this, otherwise you wildly mix values from the activation channels.

a_S = tf.transpose(tf.reshape(a_S, [n_H * n_W, n_C ]))
 a_G = tf.transpose(tf.reshape(a_G, [n_H * n_W, n_C ]))

This is because n_H and n_W come before n_C in a_S and A_G. This is crucial.
I made the same mistake in another setup.