Most examples if not all, are models that have been trained with images that are turned grey. Does this mean that models only detect edges? Why wouldnt you want to keep color so that model could learn that as well?
$\begingroup$
$\endgroup$
