Nothing I know of exists, but I've toyed with this idea before. The main trouble is mounting the camera somewhere above the board. I suppose for personal use, it isn't too big a deal.
Create app to "see" board
Sort:
You mean something like this? You could even combine it with a chess engine for a more "capable" robotic chess player, like the Raspberry Turk. There are plenty of libraries/frameworks/APIs to help you develop an app with this kind of tech, like OpenCV and Apple Vision.
Yes, exactly like that! Pair it with something like this ( https://www.instructables.com/id/Phone-Scanner-Stand/) and it's a go!
Placing the camera directly above the board shouldn't be necessary. You should be able to just point the camera towards the board from any angle as long as all the pieces are visible for the object tracker, although this might require some calibration beforehand.
Apr 13, 2019
0
#6
This requires machine learning, feeding the software with tens of thousands of hand-tagged pictures of real chess boards. There's a guy in Silicon Valley who did that, he even published his code.
Apr 14, 2019
0
#7
Something similar already exists, see https://github.com/swenae/chesscam
But I don't think it is an app for a mobile.
Apr 16, 2019
0
#9
thebioguy hat geschrieben:
No. This assumes a set up board at the beginning of the game and only tracks from which field to which the human moves a piece. It does not have to identify which piece is actually is since it knows that already. It is therefor not capable of scanning a random position on a board.
I am working on it. Ideas & funding would be greatly appreciated.
This isn't too difficult if you don't want 100% accuracy on the pieces. I did a prototype a while ago. I'm not as interested in chess currently; but if I get some time, I might wrap the thing up into something useable.
I started with wanting to add an automatic input to my analysis program where you can take a picture of a chessboard (both 3D and 2D) and get a partial FEN (the picture cannot tell you move count, castling, etc.) The program worked pretty well, getting 70%-100% accuracy depending on the image quality (e.g. computer screen and camera sensor pixel aliasing, wrapped pages in books, uneven shadowing on real chessboards, etc.).
I was working on improving each deficient area to boost the accuracy until I got distracted so the project was abandoned.
Based on my experience; I think writing a program to simply track a game's progress (like my first post in this thread) isn't hard at all and you can get 100% accuracy. All you need to do is track movements, color, promotion pieces and human hand is moving around. This seems a lot easier than what my program was doing (and that wasn't very hard).
[fixed formatting issue -- MS]
MGX just broke the Internet.
Why did he break the internet?
It seems to me that a cell phone (or anything with a camera) could be used to "see" a board and (1) record moves, (2) be used as an input device for chess computers and (3) online chess servers.
Does something like this exist??