Introduction to Neural Networks for C#, Session 13

Session Material

Optical Character Recognition (OCR) is a good application for the Self Organizing Map (SOM). To recognize any of the 26 latin characters we use 26 output neurons. These are the groups that the input letters will be broken into. The input neurons are a grid. For example, if the character resolutions were 5x7, there would be 35 input neurons. The demo program allows the user to draw the characters at a higher resolution than 5x7. The higher resolution image is downsampled to 5x7. First the image is cropped to remove whitespace. This allows the user to draw the character at any size and the program still has a good chance of recognizing the character.

Videos for this Session

Video	Title
	Introduction to Neural Networks for C#(Class 13/16, Part 1/3)
	Introduction to Neural Networks for C#(Class 13/16, Part 2/3)
	Introduction to Neural Networks for C#(Class 13/16, Part 3/3)