which "mode" is it?

I've used your code on images from MS COCO dataset.
It proposed about 400 regions per image,
which seems less than what is usually reported.
I was wondering which mode the corresponds to as it is without modification.
