model image size and input shape

Hello everyone,

I see that there are both model image size and input shape defined in the model. If we set the two parameters to the same value, then does model perform any rescaling on images or not? I also do not understand exactly grid size multiplier value which also relates input and output scale.

I need to use this model to perform some sort of visual inspection and detect some problems and cracks etc.. So, it would be better for me if it is possible to work with highest resolution. I know it would require more GPU but other than that is there a limit for max image size that we can work with?

Many thanks,