Hi, All.

I'm going to develop image segmentation engine using SOM NN.
It will be a preprocessing step of OCR.

So, I'm having an image that represents a single textline.
And I need it to be splitted into single characters.

I gave a shot to MatLab's SOM Toolbox, but the main point to me now is to understand how should it be trained.

What training data should I pass to the network?
any help appreciated.


Thanks.

If you want to train it to read text from images of text lines, you need to figure out a way to represent semantics. It's not trivial.