The goal of the present module is for it serve as a tool for
experimenting with single-instance and multi-instance object
detection and to appreciate how much more difficult the
latter problem is.  Yes, you could directly download the
code for, say, YOLOv3 (version 3 of YOLO) and have it do the
right thing on your images, but that does not serve the
purpose of education.  In order to understand in depth the
logic used in the successful object detectors, you have no
choice but to write your own code for at least the key
elements of that logic.  That's what the purpose my module
is --- to provide you with a platform for experimenting with
the key steps of the well known object detectors.

A second goal of this module related to the graph-based
method for generating region proposals.  Toward that end,
this module presents implementations for the Selective
Search (SS) algorithm for object detection as proposed by
Uijlings, van de Sande, Gevers, and Smeulders.  The
Selective Search algorithm sits on top of the graph-based
image segmentation algorithm of Felzenszwalb and
Huttenlocher (FH) whose implementation is also included in
the RPG module.

