Hi Tony, thanks for the feedback! Indeed, batch normalization is not used to do that, but as you can see in the article, I did not claim that. The part you highlighted is only explaining typical preprocesses people usually do when training neural networks.