Where exactly is normalization applied? Or is it not needed...

[davy-code]

Issue Description
I am unable to locate the specific point in the codebase where the normalization of input images occurs.

Additional Context
In my experience, image normalization is standard practice unless otherwise specified. However, I have observed improved results with the pre-trained yolov9 models when input images are not normalized. To clarify this discrepancy, I would like to understand if normalization is applied during training and, if so, in which file exactly.