AForge.NET

  :: AForge.NET Framework :: Articles :: Forums ::

OCR neuralnet need help

The forum is to discuss topics from different artificial intelligence areas, like neural networks, genetic algorithms, machine learning, etc.

OCR neuralnet need help

Postby vattic » Sat May 28, 2011 5:12 pm

hello ,

I'm working on "hand writting recognition with Neural net", i do it without sdk , i try to use the popular approach that is to save character's form in matrix ex x= [10,7] in my data base i do that for all characters : this is perfect "X"

x={ { 1, 0, 0, 0, 0, 0, 1 },
{ 1, 0, 0, 0, 0, 0, 1 },
{ 0, 1, 0, 0, 0, 1, 0 },
{ 0, 0, 1, 0, 1, 0, 0 },
{ 0, 0, 0, 1, 0, 0, 0 },
{ 0, 0, 0, 1, 0, 0, 0 },
{ 0, 0, 1, 0, 1, 0, 0 },
{ 0, 1, 0, 0, 0, 1, 0 },
{ 1, 0, 0, 0, 0, 0, 1 },
{ 1, 0, 0, 0, 0, 0, 1 },

So my net have 70 inputs, 80 nœuds hidden and 70 output. When i draw in the panel i try to get 70 pixels like a matrix this is the X manuscript recupired from the drawing box and applied pixels to the net :

x= { 1, 0, 0, 0, 0, 0, 1,
1, 0, 0, 0, 0, 1, 1,
1, 1, 0, 0, 0, 1, 1,
0, 1, 1, 0, 1, 1, 0,
0, 1, 1, 0, 1, 1, 0,
0, 0, 1, 1, 1, 0, 0 ,
0, 0, 1, 1, 1, 0, 0,
0, 1, 1, 0, 1, 1, 0,
1, 1, 0, 0, 0, 1, 1,
1, 0, 0, 0, 0, 0, 1, }

and this is the 70 output recupired from output of the net like a matrix too :

1, 0, 0, 0, 0, 0, 1
1, 0, 0, 0, 0, 0, 1
1, 1, 0, 0, 0, 1, 0
0, 0, 0, 0, 0, 0, 0
1, 0, 0, 1, 0, 0, 1
1, 1, 0, 0, 0, 0, 1
1, 0, 0, 0, 0, 0, 1
1, 1, 0, 0, 1, 0, 0
1, 1, 0, 0, 0, 1, 1
1, 0, 0, 0, 0, 0, 1

there is lot off mistakes when we campare between perfect X and the output. so my problem is how can i have decide that is X or an other character ?? i try to compare element element between the 2 matrix ( perfect X and output X), and count number off wrong pixels, finally i have a decision when i have min wrong pixels. but this idea is not effective because i have Mixing between characters the "I" regognise "T" the "B" regognise "D" ....etc i hope to give me the correct way ...
vattic
 
Posts: 4
Joined: Wed May 11, 2011 9:36 pm

Re: OCR neuralnet need help

Postby andrew.kirillov » Wed Jun 01, 2011 12:28 pm

Hello,

What are you trying to do actually? I am asking this because what you said you want to get ("hand writting recognition") does not really in sync with what you do ("So my net have 70 inputs, 80 nœuds hidden and 70 output"). What is the point of 70 outputs? What they are supposed to tell you?

In most cases, when it is required to recognize 26 characters, the output layer has 26 neurons. Maybe you are trying to do something different ... If so, then what?
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3443
Joined: Fri Jan 23, 2009 9:12 am
Location: UK




Return to Artificial Intelligence