Recognition of chess positions in pdf documents

[Dann Corbit]

I have lots of really old chess puzzle books in pdf format. Most are so old that the notation is also archaic.
I did some searching for recognition of a chess board, but everything I found seems to be relating to a computer recognizing a physical chess board as seen through a camera. I found something a decade or so ago, but it did not work well, but we have new technology now so it seems a lot more doable.

Is there any known tools for automated extraction of chess game positions from image files of book pages?

[jp]

Are your pdfs scanned pages or the originals? There's more hope for the latter, which have embedded info.

[Dann Corbit]

These books are mostly from the 1800's or so, so all are scanned.

[DrCliche]

It should be relatively simple to fine tune something like Neural Chessboard to work well on your scans.

[Pi4Chess]

Have you seen these ?

https://github.com/metterklume/chessputzer

https://github.com/IlicStefan/ChessDiagramRecognition

[Vinvin]

I tried Chessify on Android, no perfect but OK : https://chessify.me/

[Jonathan003]

You can try this webtool: https://ebook.chessvision.ai
But it is a paid service. And I don't know if it will work for old books.

[Dann Corbit]

I might try neural-chessboard, which creates uniform images first, so that I only have to train one of the others.
Thanks for all the insights.