face recognition of an image

sir i added new photos to the database . the data base is loading but its not taken in the recognition program myprogram is as follows

w=load_database();

%%Initializations
% We randomly pick an image from our database and use the rest of the
% images for training. Training is done on 399 pictues. We later
% use the randomly selectted picture to test the algorithm.

ri=round(400*rand(1,1));            % Randomly pick an index.
r=w(:,ri);                          % r contains the image we later on will use to test the algorithm
v=w(:,[1:ri-1 ri+1:end]);           % v contains the rest of the 399 images. 

N=20;                               % Number of signatures used for each image.
%%Subtracting the mean from v
O=uint8(ones(1,size(v,2))); 
m=uint8(mean(v,2));                 % m is the maen of all images.
vzm=v-uint8(single(m)*single(O));   % vzm is v with the mean removed. 

%%Calculating eignevectors of the correlation matrix
% We are picking N of the 400 eigenfaces.
L=single(vzm)'*single(vzm);
[V,D]=eig(L);
V=single(vzm)*V;
V=V(:,end:-1:end-(N-1));            % Pick the eignevectors corresponding to the 10 largest eigenvalues. 

%%Calculating the signature for each image
cv=zeros(size(v,2),N);
for i=1:size(v,2);
    cv(i,:)=single(vzm(:,i))'*V;    % Each row in cv is the signature for one image.
end

%%Recognition 
%  Now, we run the algorithm and see if we can correctly recognize the face. 
subplot(121); 
imshow(reshape(r,112,92));title('Looking for ...','FontWeight','bold','Fontsize',16,'color','red');

subplot(122);
p=r-m;                              % Subtract the mean
s=single(p)'*V;
z=[];
for i=1:size(v,2)
    z=[z,norm(cv(i,:)-s,2)];
    if(rem(i,20)==0),imshow(reshape(v(:,i),112,92)),end;

im getting error in line 2 pls help me to sort out this problem

[Andrew Newell]

I formatted your code for you. Please have a look at TUTORIAL: How to format your question.

[Andrew Newell]

What error are you getting?

[Image Analyst]

Please read this. Then paste in ALL the red text in your error message - don't leave anything out (like most people, who assume that all of us here have the Crystal Ball Toolbox).

[chaithra k n]

actually i removed all the images in the ready data base and tried to add add our photos, while loading data base i get error. this is the code for data base; function out=load_database(); % We load the database the first time we run the program.

 persistent loaded;
persistent w;
if(isempty(loaded))
    v=zeros(10304,400);
    for i=1:40
        cd(strcat('s',num2str(i)));
        for j=1:10
            a=imread(strcat(num2str(j),'.pgm'));
            v(:,(i-1)*10+j)=reshape(a,size(a,1)*size(a,2),1);
        end
        cd ..
    end
    w=uint8(v); % Convert to unsigned 8 bit numbers to save memory. 
end
loaded=1;  % Set 'loaded' to aviod loading the database again. 
out=w;

i was getting error in line v(:,(i-1)*10+j)=..... then i removed (:,(i-1)*10+j) then data base get trained but while running face recognition with my new data base its showing error in that line "size exceeds matrix dimention" .

[Image Analyst]

Try these links:

http://blogs.mathworks.com/videos/2012/07/03/debugging-in-matlab/

http://www.mathworks.com/matlabcentral/answers/13205-tutorial-how-to-format-your-question-with-markup

Answer 1

[Image Analyst]

Because you're using

ri=round(400*rand(1,1));    
r=w(:,ri);
v=w(:,[1:ri-1 ri+1:end]);

you could run into problems, like if ri = 0 or ri is more than the number of rows in w. Plus if you happened to pick the first or last column, you won't be able to calculate v.

Try this more robust way

[rows, columns] = size(w);
ri = randperm(columns, 1);
r=w(:,ri);
allColumns = 1 : columns;
% Get a list of all columns other than ri.
otherColumns = setdiff(allColumns, ri)
v = w(:, otherColumns);

[chaithra k n]

thank u sir, i corrected all the errors and now i tried to give external image, detect that image crop that image and then that croped image is used for recognition ; its working fine till croping and taking that image as input for recognition for same code .but i'm getting error in line | | p=r-m | | can anyone help me to make out the error...?

[Image Analyst]

Perhaps, if you tell us the error message. That means all the red text, not just a small snippet from it. We need everything that is in red text.

[Mandeep Singh]

These are the changes made. I picked up a photo and doing all the above operations on push of a button.

r=ke1;
v=w(:,[1:399]);

here ke1 contains the image i want to test. These are the errors.

??? Error using ==> minus
Matrix dimensions must agree.

Error in ==> face>pushbutton2_Callback at 137
p=r-m;   % Subtract the mean

[Walter Roberson]

Upgrade to R2016b. Or use bsxfun. Or repmat the mean to the same size as the data.

[Mandeep Singh]

Am I converting the image properly or not.. I am getting the image matrix of correct size but not getting the result.

global ke1;
s=zeros(10304,1);
a=imresize(ke1,[112 92]);
    s(:,1)=reshape(a,size(a,1)*size(a,2),1);
    ab=uint8(s); % Convert to unsigned 8 bit numbers to save memory. 
r=ab(:,1)                         % r contains the image we later on will use to test the algorithm
v=w(:,[1:399]);           % v contains the rest of the 399 images.

Answer 2

[Mandeep Singh]

I just want to ask that in the above program how we can take user selected image instead of selecting random images.

[Walter Roberson]

str2double the results of inputdlg prompting for the image number.

Or use uigetfile to get an image name and figure out from there which of the 39 images it was to calculate ri

[Image Analyst]

Put the list of images into a listbox where the user can simply highlight one or more. For example, see this framework: MAGIC GUI