that is so cool. i wish i could help, but i might have the problem of having 0 understanding of what u are doing exatcly.
Who wants to help train my machine learning model for FEN detection?
@
are you by any chance also an iOS user? I'm preparing a alpha build of a test app that I can invite you too which you than can use to take a lof of pictures of chessboards in books and the app will then submit those pictures and train the machine learning modelHey, @Huberand I can help you, I am currently working as a software developer, but haven't worked in Computer Vision ever, are you looking for some experienced fellows ?
@quietheathen1st are you by any chance also an iOS user? I'm preparing a alpha build of a test app that I can invite you too which you than can use to take a lof of pictures of chessboards in books and the app will then submit those pictures and train the machine learning model
No, Android.
That is a really useful tool to develop. I use chessify on android. It is ok but I expected it to capture more accurately.
@
I tried the app as well I think it works quite well. I actually don't want to release an app. I only want to write a framework that other apps can include later on. Like e.g. Chess Studio which is quite known by many chess playersI have now an alpha version ready that everyone who is interested can install under this link https://testflight.apple.com/join/Y18b7Pqh
If you are using it please use your chess.com username in the beginning as the username so that I can contact you in case I have questions regarding your scans.
Just as a disclaimer. the current scanned results might be bad but the goal of this first alpha version is to gather a lot of different images to be able to train the model better.
Thanks to everyone who is helping out.
I created a small iOS Swift library that analyzes a picture of a chessboard and renders the FEN notation of the pieces. I'm currently training my Machine learning model with a lot of different pictures of 2d top-down chessboard photos. If you are interested in helping me feel free to reach out to me. I'm looking for sharp photos taken from books or newspaper chessboard puzzles to train my model even more.
Below you see how its currently working. I feed my framework with a picture of a chessboard in a book and then it analyzes the picture, finds all the squares, and passed it to the machine learning model which then analyzes each square to predict which chess piece is on the square.
Analyzed position from the above image.