Maplocate implements a new methodology to automated map georeferencing based on toponym recognition and pattern matching algorithms. This methodology is publicly available and described in more detail at:
Karim Bahgat, and Daniel Runfola. Toponym-assisted map georeferencing: Evaluating the use of toponyms for the digitization of map collections. PLoS One. 2021 Nov 18;16(11):e0260039. doi: 10.1371/journal.pone.0260039.
All code and other software developed as part of the toponym-based georeferencing methodology has been open-sourced and can be used by researchers and developers to implement or improve on the methodology:
Python package with various tools and utilities optimized for optical character recognition (OCR) on map images.
Python package for detecting toponyms from map images and matching these to their most likely geographic coordinates. Leverages Mapocr for text recognition.
Python package for efficiently fitting and predicting 2D coordinates based on various image and geographic coordinate transformations. Used for estimating the transform function based on the identified toponym coordinates and creating the final georeferenced image.