Training a support vector machine in C# for character recognition

Question

0.00/5 (No votes)

See more:

, +

For those new to SVMs, what would be the best place to start if they want to train a SVM in C# on different images of characters distributed in their respective folders?

This is an assignment in our Image Processing course. We have been doing different IP algorithms on images in C# but, with almost no help from instructor, I have no clue how to deal with training SVMs in C# on a bunch of character images.

We haven't studied Machine Learning course yet. And through googling I have gotten some basic knowledge of SVMs. But still have no idea how to implement for training data in C#. Is there any library? As per my knowledge the machine will output a training file that will be used to recognize the text from images.

The training data looks like this: http://www.tiikoni.com/tis/view/?id=6b26c8e

Any help is appreciated.

What I have tried:

Tried understanding the basics of SVM. Not sure where to start.

Posted 17-May-16 8:42am

AshRayX

Updated 22-Oct-18 1:07am

Add a Solution

1 solution

Add a Solution

Add your solution here

Treat my content as plain text, not as HTML

Preview 0

…

Existing Members

Sign in to your account

...or Join us

Download, Vote, Comment, Publish.

Your Email
Password
Forgot your password?

Your Email
This email is in use. Do you need your password?
Optional Password

I have read and agree to the Terms of Service and Privacy Policy
Please subscribe me to the CodeProject newsletters

When answering a question please:

Read the question carefully.
Understand that English isn't everyone's first language so be lenient of bad spelling and grammar.
If a question is poorly phrased then either ask for clarification, ignore it, or edit the question and fix the problem. Insults are not welcome.
Don't tell someone to read the manual. Chances are they have and don't get it. Provide an answer or move on to the next question.

Let's work to help developers, not make them feel stupid.

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

Matt T Heffron · Accepted Answer · 2016-05-17T09:18:00

Solution 1

Yes there are SVM libraries.
Bing search of "libsvm csharp" gives over 10,000 results (how many are useful?).
There are SVM implementations for use in .NET applications.
I suspect there are examples on using them.

If you really want to learn about Machine Learning and Classification systems, check out the Free, introductory Machine Learning online course from Caltech![^]
It won't give you the code, but you'll learn lots about how to do it yourself!
If you try taking the course, I suggest using Octave (a free Matlab-compatible system)[^] for practicing and homework.
SVM is covered in this course.
You probably want some of the ML background before digging into SVM. There's a reason it isn't near the beginning of the telecourse from Caltech! (My alma mater!)
Good luck.

Posted 17-May-16 9:18am

Matt T Heffron

Comments

Sergey Alexandrovich Kryukov 17-May-16 15:40pm

5ed.
—SA

Matt T Heffron 17-May-16 16:50pm

Thanks

AshRayX 17-May-16 16:07pm

Thank you for replying. I understand ML background is required for a proper understanding. Since time is short, I don't think I will be able to take this course before the submission. As a beginner I want to know this: Would the libraries like libsvm or opencv etc. have functions that would take the training data to get a training file OR do we have to convert the training image to data to mathematical model/representation first?

Matt T Heffron 17-May-16 17:43pm

Following up on my comment below.
ML on raw images is probably computationally intractable.
(At least, a very large computational requirement.)
Determining a small feature set that can be pre-computed easily is probably the only practical strategy.
My first thought would be first to trim and scale all the images to the same size (at least approximately align the baseline and horizontally center the trimmed image in the final size "canvas"), then get some simple features like horizontal and vertical symmetry, overall intensity.
Then, you should split the feature data into training and testing data sets.
Do this split randomly! Do not peek when choosing which data are for training vs. testing!

AshRayX 17-May-16 16:33pm

Also what would be the most suitable library for this particular case. There is EMGU CV (OpenCV wrapper), libsvm C#. Whats the most suitable one? Any other one out there?

Matt T Heffron 17-May-16 17:04pm

When I took the course in Nov. 2012, I used libsvm from Octave and only used what was necessary for the SVM homework. I don't have recollection of the details of libsvm capabilities and usage model. Neither have I used EMGU CV, libsvm-csharp, etc.
Here are a couple of additional links to SVM/C# related articles (using a different free ML library, Accord.NET):
http://www.codeproject.com/Articles/106583/Handwriting-Recognition-Revisited-Kernel-Support-V
http://crsouza.com/2010/04/kernel-support-vector-machines-for-classification-and-regression-in-c/
Sorry I can't offer more assistance.