I have a lot of 1-band and 3-band raster layers which I will be using as input for creating training samples (the training samples will be used to train deep learning models). My question is, if I modify the symbology i.e. colour scheme, stretch type etc, will these modifications influence the training samples created using these raster layers?
It is my understanding symbology is largely an aesthetic feature. I believe that when using raster layers to create training samples, the information being used is the raw values of each pixel, which as far as I know is not changed by any aspect of the Symbology tab. Therefore, in theory, symbology should not affect the training samples. But I was hoping someone else could confirm it for me.
