03-29-2025, 04:30 PM
I am very happy for you to set the spec for the X O detection. You can decide the grid size in pixels, and how many samples we will create for each other to test with - this is so that neither of us can do the training with the data that will be used for the test.
So for example we could decide on a 10x10 grid, train out AI in any way we like, and each generate 20 test cases which we used at least as part of our training and our AIs can handle. At this point we can swap our 20 test cases and see what happens.
It will be tricky to create the training and test cases if not made by hand - part of the fun in this will be creating the training set - I would probably write a program to do it - maybe from lots of different fonts positioned and scaled randomly and maybe some hand drawn and mayer other methods...
So for example we could decide on a 10x10 grid, train out AI in any way we like, and each generate 20 test cases which we used at least as part of our training and our AIs can handle. At this point we can swap our 20 test cases and see what happens.
It will be tricky to create the training and test cases if not made by hand - part of the fun in this will be creating the training set - I would probably write a program to do it - maybe from lots of different fonts positioned and scaled randomly and maybe some hand drawn and mayer other methods...