Adding image-splitter that uses Canny-Edge detection.

[iamgroot42]

Addresses issue #1

Using the fact that pages are more-or-less of the same size, and that there is some space between them, I have to build a method to split such given images into two pages. This method does not require any training or annotations. Based on the assumption that the images won't be too rotated, it will give good results.
Some visualizations from the given sample dataset are:

Left half:

Right half:

This method is quite fast: it splits all the given images in a matter of seconds and does not require high-end GPUs or deep machine-learning models (or humongous data) to train it.