a better code for LDA based face recognition

I have modified the former code so that it runs faster and can be run on normal PCs.

Description

This code is LDA base face recognition programme. It reads nots(here 5) faces from ORL database and the rest (noc-nots) are used as test. LDA_Performance shows the recognition performance.
   
In order to be able to run this programme for ORL face database you need to download the following zip-file. Then copy this code in the folder which includes the face samples and then run this script. The zip-file is available at: http://www.mathworks.com/matlabcentral/fileexchange/22466
   
This code has been written in Fall 2008 by me, Gholamreza Anbarjafari (Shahab). You can use this code for any research and academic purposes and you may refer to us in your acknowledgement!
   
Feel free to contact us for any further information:
 {hasan.demirel, shahab.jafari}[AT]emu.edu.tr
  http://faraday.ee.emu.edu.tr/shahab
  http://faraday.ee.emu.edu.tr/hdemirel
  (c) Demirel and Anbarjafari -2009

• Required Products: Image Processing Toolbox
• MATLAB release: MATLAB 7.6 (R2008a)

[Anon]

When this script is "better", why not update the existing LDA file? Have you never heard of the update function?

This file needs major revision. As it stands right now, it is useless. It is a script and not a function

clc
clear all
close all
clear memory

It reads an image that is not supplied with this function.

And again: Why do you repeatedly clear the variable memory that does not exist. Or does the command clear memory make sense? Correct me if I am wrong. As it stands right now, the script is only useful to you. Since there is no help, it does not even have educational purpose.

[Shahab Anbarjafari]

Plus, it is called better coz now you can run the code on simpller computers (maybe like the one that we usually use, rather than super fast computers).
Best
Sh

[John D'Errico]

First of all, please DON"T rate your own code. Your opinion is not pertinent here. We know you are impressed with your own work, otherwise you would not have put it here for the world to see.

PLEASE do not continue to resubmit the same file as new work however. This just clogs up the FEX, leaving too much crap out there to wade through.

As for your work being of any true value, I won't say too much, except that this is a malicious script file that clears the user's workspace. It reads a file that is not even supplied, so the code does not work as is to test it.

The script has NO help. NO explanation of what it does. It uses loops to do trivial things that do not need loops.

[Shahab Anbarjafari]

:)

[Omid Khanmohamadi]

This submission can serve as an epitome of the unfortunate trend of the scientific world today. Lots of people are looking for quick, black-box solutions to sophisticated problems. (How totally far that is from being of any value is a separate issue.) Any work (this submission included) containing some buzzwords will attract lots of attention without necessarily being of any quality. I won't repeat John's comments regarding the low quality of this work. I add that apart from rating his own work, the author mentions on his website that he "has been ranked 1 by MathWorks since 19th February 2009." He forgets to mention, however, that, first of all, this ranking is not done by The MathWorks and that it is based on the number of submissions and downloads, and not on the user ratings which could be the true indicator of the quality of a work, based on which this work (and similar works of this author) are of no quality - an average rating of 2.2/5 based on 21 ratings, as of today.

[Shahab Anbarjafari]

:) may God help you and you improve your position from 16512 to 16000 . Thanks for comment!

[Xu Wings]

Apart from couple of comments from some amateur users I have to appreciate your work. Very useful code of LDA face recognitions. I also appreciate your work in PCA. Helped me a lot. Both rank 5.

[John D'Errico]

Actually, the number 1 ranking of this author on the website is a farce anyway. Someone has been cooking the books, for "fun", or what ever reason. So those rankings are totally random. See that of the top 100 authors on the website, each individual has roughly one file with a relatively huge number of downloads. The number is inappropriately large for the actual file chosen.

In fact, these spurious downloads form well over half the total downloads on the website. I once calculated that one individual was artificially generating close to 75% of the download traffic on the website. I've also made sure the website administrator knows of these facts. My suggestion is they cut off the IP address completely for the phantom downloader, as it likely represents a large waste of their resources and a drag on the website response.

So this author should not be so proud of his #1 ranking. His 15 minutes of fame will finish soon.

[Shahab Anbarjafari]

:)

[Yi Cao]

I just tested this code with the face database provided in another submission. The code does not work at all. It reports several errors. For example:

??? Undefined function or variable 'x'.

Error in ==> ldaface at 53
AVERAGE=mean(x')';

I corrected this error by rename x to X. Then I got another error:

??? Undefined function or variable 'noc'.

Error in ==> ldaface at 54
Average_Matrix=(ones(noc*nots,1)*AVERAGE')';

I then follow the PCA code to define noc=40; Then, the program finishes with

LDA_Performance =

     0

Clearly, this code has not been tested before submission. I doubt how this code could be rated by someone as useful.

[Shahab Anbarjafari]

Thanks Xu

[Dimitri Shvorob]

 

[Dimitri Shvorob]

Typical loop-heavy novice script displaying unfamiliarity with the syntax of basic commands like 'max', 'find', 'mean'. I am sure this simple textbook calculation has shown up on FEX before, and better, so look around.

[Kemal]

 

[Hal 9000]

poorly written

[Ali]

 

[Ziba]

Thank you

[V. Poor]

 

[alain boyer]

 

[Xing Yu]

 

[Robin Torres]

 

[Mathew Sermen]

really excellent work

[Mustafa]

Just what I was looking for!

[Xu Wings]

 

[Nikola]

Very helpful

[Xu Wings]

 

[Mak K]

first of all, I couldn't run the file because it doesn't have the files it needs

I read the code it is not LDA, it's PCA because you did the following:
1- loaded the pictures for training in one matrix as rows (face space)
2- loaded the rest of the pictures for test in another matrix
3- subtracted mean face from all of face space (zero mean face space)
4- covariance for zero mean face space
5- eigenvectors for covariance matrix
6- project the eigenvectors on every picture in train and test sets
7-measure the euclidean distance for test set
 
that is PCA

for LDA, you must get mean for each class
then the mean of all the classes
calculate within-class scatter matrix
calculate the between-class scatter matrix
then the eigenvectors for the product and so on

which you didn't do

[Jesus Callirgos]

The file 22466 doesn't exist , please updates the information, I´m very insterted in this work,it could be helpfull, but I don`t understand so much.

Thank if you can explain me.

[asha rani]

after reading cmments, I don,t know I should use it or not!