Does Tensorflow normalize input data by default?

Question

Does anyone know if Tensorflow normalizes input data by default?

I have grayscale images that values range from roughly 20000-28000. When I normalized the data something odd happened. The network trained for a couple of hundred iterations was doing well in terms of making accurate predictions, but suddenly all predictions went to NaN. Of course, it couldn't recover because TF can't optimize from NaN.

When I didn't normalize the data training went fine and converged.

Any ideas?

Answer 1

The answer to your question is no, TensorFlow doesn’t normalize your input data by default. You have to apply normalization function in your layer like as follows: 

x_train=tf.keras.utils.normalize(x_train, axis=1)

This image is before normalization:

And after normalization, it looks like this: