Click here to Skip to main content
15,936,903 members
Please Sign up or sign in to vote.
1.50/5 (2 votes)
Hello all,
I' m Hendy. I'm newbie in neural networks I was trying to develop some ocr project and implement backpropagation method. I have been already sets the train function, but I still confusing about using the output result. How can I use the output to recognize some character? Do I need to make comparison to the binary of some character? Here is the following steps that I have been do:
TRAINNING
1. Converting text input to image
2. Normalization character image size into 10x12 matrix
3. using backpropagation.
4. get the weight and saved it.
RECOGNITION
1. Preprocessing input image(scanned text document)
2. Character image segmentation
3. Character image size normalization without thinning the character(I don't use it)
4. Convert the segmented and normalized character image into binary and set it into array (10x12 matrix).
5. backpropagation (forward propagation)
6. get the output

I have no idea to use the output to recognizing character?
Sorry about my bad english :).

Thanks everybody!
Posted
Comments
Sergey Alexandrovich Kryukov 6-Nov-12 10:26am    
Your question is clear enough, but I don't think it's suitable for this forum. The problem is not so simple and requires a lot of learning and work. One question cannot get you into it.
--SA
Hendypodz 7-Nov-12 4:50am    
Ok. Thank you.
Tristan32 11-Nov-12 0:23am    
Well I know you can make bitmaps of font characters, and then search for them inside images. Search for them rotated, and different sizes. But as to the best way to go about all of this I have no idea. You'd probably have to use glyphs, some way to load a ton of onts/characters, and search for them onscreen at a fast rate.

This content, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)



CodeProject, 20 Bay Street, 11th Floor Toronto, Ontario, Canada M5J 2N8 +1 (416) 849-8900