This tutorial will quickly demonstrate how to use Recog to recognize mouse-drawn characters. Before starting, verify that the option None is selected in the Learning menu.
The first step is to load the appropriate character network. Go to the File menu and choose Open Character Net. Open the file called VOWELS.RCG. It contains the neural network needed to recognize the capital vowels A, E, I, O, U and Y.
Next, use the mouse to draw one of the vowels. Draw it in the large empty square underneath the label no picture file. The left mouse button draws, the right mouse button erases, and double-clicking clears the picture.
When you have finished drawing your vowel, click the Digitize button. This will scan your picture and convert it to a grid of 48 pixels that the neural networks can interpret. When it is done, click the Recognize button.
Recog will immediately tell you the probability that the character you drew was an A, an E, and so on. This is called a competitive algorithm: each possible vowel is comparing itself to what you drew, and the one the matches the best is the winner. Remember that this network was trained on my handwriting, so even though it will probably pick the correct letter, the percentage may not be spectacularly high.
Try drawing, digitizing and recognizing several other vowels. Try inputting an H and see what it thinks. To train it to recognize new characters, try the next tutorial.