[Computer-go] I need an off-the-shelf final position live/dead evaluator
robertfinkng555 at o2.co.uk
Sat Nov 27 16:26:49 PST 2010
It strikes me that the extraction of the board position from an image is
less well researched than scoring a known position? I have a colleague
who is a computer vision expert (and a go player) who knocked up a
program to record a game of go by observing the board. He found that the
vision side of it was harder than he'd anticipated, despite the
apparently constrained nature of the problem.
Hope the below are useful:
On 26/11/2010 18:43, Jim Babcock wrote:
> I'm working on Go Scoring Camera, an Android cell phone app that will
> photograph a board, do image processing to figure out where the stones
> are, and score the game for you. For the scoring part, I need a liveness
> evaluator; and since this is a previously solved problem, I figure it
> would be best to take one off the shelf. So, I need a recommendation.
> I need an evaluator that will take a final position and mark which groups
> are live and which are dead.
> * Preferably in Java, but C or C++ would also work
> * If it's embedded in a larger context (such as an AI), it should be
> easy to separate.
> * The hardware budget is about five seconds of CPU time and 2MB of
> * It doesn't have to be perfectly accurate (the output will be inspected
> by a human), but higher accuracy is strongly preferred
> Finally, I need the source code under a license such that I can
> incorporate it into Go Scoring Camera, which will be payware, so no GPLed
> programs unless the original author provides me with a separate license. I
> am willing to pay for a license, if necessary.
> If you know of or have written a position evaluator and you'd like me to
> consider it, reply on this mailing list or email me directly. Please mention
> what language it's written in, a brief summary of the algorithms it uses,
> an estimate of its memory usage, and the license terms it's under.
More information about the Computer-go