Sudoku Scanner
A few months back I was riding the train home from the Studio. I had recently finished a big writing push on my book and so I finally had some free time on the train again. So I just sat there for a while doing nothing (not as easy as it sounds). That’s when I noticed a woman playing Sudoku. I considered passing the time by playing Sudoku myself. And then I remembered writing an unbeatable Tic-Tac-Toe robot and how much more fun it was than actually playing Tic Tac Toe. So I got it into my head that I wanted to write a Sudoku solver. So I did. I put it up on GitHub for all to see. And then Jim Weirich did the same. And then Corey Haines did too.
For the next couple weeks I was a obsessed with the solver and was always blabbing about it at geekfest. And that’s where the idea of a Sudoku Scanner was born. I decided then and there that I would write some software that could comprehend a photo of a Sudoku board. I burned the midnight oil on more than a few nights researching the domain, stumbling through problems, and rewriting solutions, but eventually I got my 4 test photos scanned perfectly! And I’m excited to announce that in the upcoming weeks, we will be launching a web-based version of the Scanner which should help us get more diverse test data to improve its accuracy.
I’ll probably do another rewrite in the coming months thanks to an excellent Neural Network lesson from Aslak Hellesøy at speakerconf. I’ve also received a ton of excellent ideas, feedback, and support from my fellow Obtivians, particularly Nate Jackson and Tyler Jennings. And I can’t not mention my friend and customer, Gary Levitt who generously developed and donated the Sudoku Scanner website’s design.
