A simple FSM Framework and example application written in C++, based on ideas presented in “Programming Game AI by Example” by Mat Buckland.
Download: source
In artificial intelligence, SOMs, or Self-Organizing Maps, are a type of neural network. They can be used to ‘teach’ a computer to recognize various attributes. This project deals with two demos: SOM_Color and SOM_Image. Both are written in C#.
SOM_Color is a color recognition program. A grid of nodes is initialized with random colors. The application learns how the colors are related to each other, and will sort the colors so that like-colors end up near each other. Color “similarity” is defined by their respective RGB color vector Euclidean distance.
SOM_Image is an image recognition program. A database of images is presented to the application (Madden NFL ’06 images are provided with the download). The program will group similar images near each other. Two classifications are used for this similarity calculation: a histogram of lights/darks and colors. This application works surprisingly well.
There is also a paper that goes along with this project, to better explain the utilized techniques.
Download: source