I'm playing a game where our Guild is constantly sharing teams as screenshots in Discord. I'm wondering if I can find something that will convert those team images into text, using a library of exact images for each of the troops.
Online OCR does a marginal job because of the backgrounds behind the text, so it's not a great solution. Also, it basically solves the wrong problem.
I've found mention of the OpenCV library (for Python?), but the GitHub projects I've seen are looking at a much broader problem.
Here's the algorithm I have in mind:
Locate the team in the image and resize to a standard size.
Locate the sub-images for each troop.
Search the library of template images (about 800 -- a big job to create!) for each troop and find a sufficiently good match.
Look up and return the text associated with each match.
Ideally, I would like to point the software to an image file and it would tell me the names of the troops in copyable text.
This might be an interesting enough project for me to finally install and learn Python (but that's a whole other question). For now, let's see what already exists that might work.