“A big, dumb, balding North American ape with no chin and a short temper.”
Kwyjibo is also an experimental, automated referee for Scrabble games. It was built to explore real-time OCR, computer vision and machine learning as the subject of my dissertation at the University of Sheffield in 2011.
You can read a little bit about how it works in this explanation. Much of the image processing is done using the awesome Aforge.NET computer vision library. There's also a video of it working on shefcompsci's channel.
Since this was just a research project while I was an undergraduate, it's obviously not anything ground breaking. That also means the code is a bit ugly and doesn't always follow best-practices due to time constraints. I don't usually like releasing too much rough code, but I don't mind since a few interested Scrabble players requested it.
It may not work with your own board and camera, but feel free to give it a go, if it works give me a shout!
My implementation of the Scrabble game rules is a little shaky (it's a worse referee than Homer). I'm not really a player and I didn't have much time for that bit, but the computer vision part works nicely.
Read more about this thing here in my post.
Note: this project was for academic research only and is not associated with Scrabble ®
Try it out
If you just want to try it out, you'll need a web cam, download the executable from the Kwyjibo page and run it. Follow the instructions shown to calibrate it to your board and camera.
You can also poke around with the source code on github, you'll need Visual Studio installed to build it.
Kwyjibo is licensed under The MIT License (MIT)
Copyright (c) 2014 Liam Brummitt
This license is also supplied with the release and source code. As stated in the license, absolutely no warranty is provided.