- Blind mode tutorial
lichess.org
Donate

Converting images of chess diagrams to FEN

Good morning lichess comrades,

I wrote a program that can recognize positions from books and magazines and give you the FEN with a lichess link. Here is the web interface:

https://www.ocf.io/abhishek/putz/
(Chessputzer )

You can upload the image or paste directly from the clipboard.

Note: it is designed for the the most common fonts used in print. It won't work for images from Chessbase etc. For that, use Fenbot.

My aim was to make it good enough to read blurry images from old chess books so it should be pretty tolerant of artifacts and such.

Still a work in progress so caveat emptor :)

Good morning lichess comrades, I wrote a program that can recognize positions from books and magazines and give you the FEN with a lichess link. Here is the web interface: https://www.ocf.io/abhishek/putz/ (Chessputzer ) You can upload the image or paste directly from the clipboard. Note: it is designed for the the most common fonts used in print. It won't work for images from Chessbase etc. For that, use Fenbot. My aim was to make it good enough to read blurry images from old chess books so it should be pretty tolerant of artifacts and such. Still a work in progress so caveat emptor :)

Hi scuffi,
Right now it is only designed for images from books, since that's what I used it for.

Example (click on the random example link for more):
https://cdn.pbrd.co/images/GTLRmSD.png

Hi scuffi, Right now it is *only* designed for images from books, since that's what I used it for. Example (click on the random example link for more): https://cdn.pbrd.co/images/GTLRmSD.png

I read that but I thought that if your algorithm can identify pieces of a quality that low as in your examples, the high quality of my screenshot should be no problem.

Still a very cool project :)

I read that but I thought that if your algorithm can identify pieces of a quality that low as in your examples, the high quality of my screenshot should be no problem. Still a very cool project :)

Thanks.

Right now I use a pretty low tech algorithm which is just about sensitive enough to distinguish between the black and white queens in my example.

Unfortunately the price of recognizing more fonts is more errors. For example, a pawn in one font might be too close to a blurry bishop in another.

By the way, I also reject outcomes have more than two queens etc. and are absent from 99.99% of games. Another trade off :)

Thanks. Right now I use a pretty low tech algorithm which is just about sensitive enough to distinguish between the black and white queens in my example. Unfortunately the price of recognizing more fonts is more errors. For example, a pawn in one font might be too close to a blurry bishop in another. By the way, I also reject outcomes have more than two queens etc. and are absent from 99.99% of games. Another trade off :)

This topic has been archived and can no longer be replied to.