Math symbol image recognizer used by Min image recognition server
Matlab
Latest commit 75aa434 Mar 14, 2014 @sxz8564 sxz8564 image recognizer
image recognizer used in min
Permalink
Failed to load latest commit information.
cc_license
src
AUTHORS.txt
LICENSE
README
gridVecAs.mat
symbolCodeAs.mat

README

--------------------------------------------------------------------------
  DPRL ImageRecognition

  Copyright (c) 2012-2013, Siyu Zhu, Lei Hu, Richard Zanibbi
  *Contact information is provided at the bottom of the file.
--------------------------------------------------------------------------

:: License :: 

	These tools are provided in the framework of ‘Min’ math equation recognition 
	interface 'as is' without any guarantee of suitability for non-research
	use.  No commercial use is permitted. The tools are being distributed under
	a Creative Commons license (please see the LICENSE file, and the directory
	cc_license contains a file detailing the specifics of the license).

:: Requirements ::

	MATLAB (7.0.11.584) are required to run
	the tools included.  Scripts are written in MATLAB.

----------------------------------
  INSTALLATION  
----------------------------------

No installation is needed, just copy and paste the code to the directory where you want to use the function.

----------------------------
  Training Samples (MAT files)
----------------------------

Training samples are stored in .mat files and could be loaded by Initializaiton.m
They include GridVecAs.mat and symbolCodeAs.mat which contains training samples and
corresponding symbol labels respectively.

----------------------------
  Scripts 
----------------------------

Scripts in src/ are used to accomplish the feature extraction, image recognition 
and label conversion. All programs are written in MATLAB (7.0.11.584 compatible). 

Initialization
	initilizing the system by loading training data feature vectors and
	class labels; Open MATLAB Pool for parallel computing

ContourExt
	Extract contours from images 'cc1.tiff', 'cc2.tiff', etc. The
	bounding-box information is provided in 'inputFileName'{'bb.csv'} and contour
	coordinates are stored in file 'outputFileName'{'contour.csv'}.

Code2Class
	this function is used to convert the hexadecimal infty code to its
	corresponding symbol class InputSymbolCode is not a number but a string

Recognizer
	Generate isolated symbol image from input image and bounding box 
	information (output of Segmenter). Recognize symbol class using Nearest 
	Neighbor method. Output is ICDAR symbol code. 

--------------------------------
  Contact
--------------------------------

Siyu Zhu
sxz8564@rit.edu
Rochester Institute of Technology, USA

Lei Hu
lei.hu@rit.edu
Rochester Institute of Technology, USA

Richard Zanibbi
rlaz@cs.rit.edu
Rochester Institute of Technology, USA