Add the basic project layout and minimal functionality

[wrznr]

This is supposed to be an OCR-D processor which someday will give plausibility feedback on a page's segmentation. It uses https://pypi.org/project/Shapely/ as proposed by @bertsky.

[bertsky]

Thanks for breathing life into this!

Does this minimal functionality reflect a common/specific use case you observed on some pipeline, or is it rather exemplary?

BTW, I found a good explanation of intersects vs overlaps vs crosses in the manual, which is much more than in the docstring. It does not say anything on covers though.

With the second commit, I can already see this module as going into the more general "layout alignment" (by aligning with itself) – maybe we can ultimately make 2 different processors (sharing the same heuristics), one for evaluation, one for alignment/merging?

[wrznr]

@bertsky Could you pls. accept my invitation as a collaborator? Would like to assign you as a reviewer.

2 different processors

I do not like the idea. Analytics (and this implementations) would highly overlap. I'd like to control analyze vs. modify via parameters.

This is one of the intended use cases. Happens quite often with tesseract and has a very negative influence on resegment/clip (and ultimately text recognition).

[bertsky]

Could you pls. accept my invitation as a collaborator? Would like to assign you as a reviewer.

Found it.

2 different processors

I do not like the idea. Analytics (and this implementations) would highly overlap. I'd like to control analyze vs. modify via parameters.

Yes, I guess there will be too much overlap.

This is one of the intended use cases. Happens quite often with tesseract and has a very negative influence on resegment/clip (and ultimately text recognition).

Thanks, that's very illustrative! True, this would not work at all with resegment/clip (the rules of which I rather optimised in view of present GT).

[bertsky]

This could be very useful for GT as well (example from buerger_gedichte_1178 page 0002):