AForge.NET

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

how to recognize a character from a pic with neural network

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

how to recognize a character from a pic with neural network

Postby sleepy_hollow » Tue Apr 12, 2011 12:27 pm

hi,

i want to create an OCR with neural network with backpropagation learning with the inputs are a binary image with static pixel (ex. 8x10)

so...how must i do first???
sleepy_hollow
 
Posts: 23
Joined: Fri Jan 21, 2011 12:40 pm
Location: indonesia

Re: how to recognize a character from a pic with neural netw

Postby andrew.kirillov » Tue Apr 12, 2011 7:01 pm

Hello,

You may find some info here (see the "Popular approach" chapter). If you google a bit you will find a lot similar.
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: how to recognize a character from a pic with neural netw

Postby sleepy_hollow » Wed Apr 13, 2011 1:35 am

in this topic (viewtopic.php?f=5&t=1864) you suggest 1 and -1 for bipolar sigmoid function and in your link that suggest me you use 0,5 and -0,5. so which one should i use

after training a neural network, how to save that training and what actually i should save from that training
sleepy_hollow
 
Posts: 23
Joined: Fri Jan 21, 2011 12:40 pm
Location: indonesia

Re: how to recognize a character from a pic with neural netw

Postby andrew.kirillov » Wed Apr 13, 2011 8:19 am

sleepy_hollow wrote:in this topic (viewtopic.php?f=5&t=1864) you suggest 1 and -1 for bipolar sigmoid function and in your link that suggest me you use 0,5 and -0,5. so which one should i use

Both should work fine. However, you are welcome to try and experiment.

sleepy_hollow wrote:after training a neural network, how to save that training and what actually i should save from that training

Did you check docs? There is Save() method for neural network class - it saves weights of neurons (all you need).
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: how to recognize a character from a pic with neural netw

Postby sleepy_hollow » Wed Apr 13, 2011 10:15 am

Did you check docs? There is Save() method for neural network class - it saves weights of neurons (all you need).


i barely know about that, and i had check it out
still i don't get it how to save it, what data might be and where it'll be saved
and in your neuro OCR (from codeproject) i don't find method to save result from training (may be i'm not properly check it)

please explain or give an example if it doesn't bother you

thanks before
sleepy_hollow
 
Posts: 23
Joined: Fri Jan 21, 2011 12:40 pm
Location: indonesia

Re: how to recognize a character from a pic with neural netw

Postby andrew.kirillov » Wed Apr 13, 2011 10:36 am

sleepy_hollow wrote:and in your neuro OCR (from codeproject) i don't find method to save result from training (may be i'm not properly check it)

That code is old, not updated for a while and no longer supported - it is just left there for users to struggle with it on their own (just read the article + code).

sleepy_hollow wrote:please explain or give an example if it doesn't bother you

See the Save() method.

sleepy_hollow wrote:and where it'll be saved

To file or to stream (there are two versions of Save() method).
With best regards,
Andrew


Interested in supporting AForge.NET Framework?
User avatar
andrew.kirillov
Site Admin, AForge.NET Developer
 
Posts: 3436
Joined: Fri Jan 23, 2009 9:12 am
Location: UK

Re: how to recognize a character from a pic with neural netw

Postby sleepy_hollow » Wed Apr 13, 2011 3:02 pm

so if i use the string method, is it gonna be saved in a file?
and what the difference with stream method?
sleepy_hollow
 
Posts: 23
Joined: Fri Jan 21, 2011 12:40 pm
Location: indonesia




Return to Artificial Intelligence