|
|
|
|
Object and Pattern Recognition
I want to extend my pattern recognition algorithms that I used within my paper rock scissors ai to either more general
cases of pattern finding or to a more complex game. The paper rock scissors ai was primitive yet effective because
the game is simple, more complex pattern recognition would be required for a game such as chess or go.
I also want to eventually attempt an object recognition system. Currently, I don't know what method I think should work.
There are several research groups currently working on Object Recognition systems using conventional and some less
conventional methods. The key to all the systems is pattern recognition, and that I feel is the flaw. Obviously, our
brain utilizes pattern recognition when receiving input from our eyes to determine what we are seeing, but there in my
opinion is more too it than just visual pattern recognition.
As a small thought experiment imagine a program that uses pattern recognition to make accurate predictions. The system
will never be 100% certain on its next prediction (unless of course the pattern is very simple). Because of this all
pattern recognition systems offer only the most PROBABLE prediction which turns out wrong sometimes. If you have
downloaded my paper rock scissors AI you know what I mean. The algorithms are pretty accurate but sometimes they are
wrong. However, when was the last time you walked into a room, saw a futuristic chair you had never seen before that
was not occupied, and you mistook it for something else? Chairs and any human made item can come in any size, shape,
texture, color and yet flawlessly and without prior knowledge for the most part we recognize it immediately. Possibly
aside from visual pattern recognition there is functionality pattern recognition occuring in the brain as well. Whatever
it is, I believe only limited success will come from strictly visual pattern recognition systems.
|
|
|
|
