Error using * Inner matrix dimensions must agree.

function [  ] = Eigenfaces( ) 
z=[];
 ta=[];
 for i=1    
 str=int2str(i);   
  str=strcat(str,'.JPG'); 
    im=imread(str);   
im=rgb2gray(im);
  im=imresize(im,[256 256]);  
   temp=reshape(im,65536,1);  
   z=[z temp];m=mean(z,2); end     
 for i=1:size(z,2)        
 t=double(z(:,i))-m;       
  ta=[ta t]; 
  end    
 R=ta'*ta;    
 [v,d]=eig(R);   
  L_eig_vec=[];   
 for i=1:size(v,2)      
  if(d(i,i)>1)      
      L_eig_vec=[ L_eig_vec ;  v(:,i) ];   
     end   
 end    
 Eigenfaces = ta*L_eig_vec;
     disp('eig finished');
 projectedimages = []; 
Train_Number = size(ta,2);
 for i = 1 : Train_Number 
temp = Eigenfaces'*ta(:,i); he
 projectedimages = [projectedimages temp];
end
target=eye(Train_Number);
save latest.mat
end

Answer 1

[Sulaymon Eshkabilov]

Hi,

There is one very tiny but a really crucial point missed with the index of z(). Here is corrected part:

...    
z=[z temp];m=mean(z,2);
end
for i=1:size(z,2)
    t=double(z(i,:))-m;   % index of z(:,i) has to be z(i,:). That is all
    ta=[ta t];
end

Good luck

[Bereket Ayalew]

thank you very much!!!!!!! it starts working normally after these change............ but i do have also another question and that is...... i want to changes many faces in the same directory to train in neural network..... so i use the for loop

for i=1:100

like this to change the images iteratively upto 100..... then it says error in this code line

t=double(z(:,i))-m;

saying "error using '-' sign".... i already rename them like 1 2 3 4 5..... then put them in the correct directory...... but it say also "error in the dimension" on the code line i mentioned above!!! can you help me with this?? i really want them to change to eigen and in the same folder means latest!!!!!!!! thank you for your cooperation!!

[Greg Heath]

Please reformat

Greg

[Bereket Ayalew]

what does reformat mean?? what thing is i reformat??

Answer 2

[Sulaymon Eshkabilov]

Hi,

I could not find any problem with your corrected for a few .jpg images (created) to process that I've played with.

There was another minor typo (see below given): temp = Eigenfaces'*ta(:,i); %he

The only thing might be the casue I think is that the file extention has to be unique in all of your files, e.g. jpg or jpeg (if mixed then fix them). Noe that your code works with *.jpg only! if you would use *.jpeg then correct the code for *.jpeg extension. I could not generate your problem with my MATLAB. I've tested it with 2017a and 2019a so far. Btw, what version of MATLAB are you using?

function [  ] = Eigenfaces( ) 
z=[];
ta=[];
for ii=1:100
    str=int2str(ii);
    str=strcat(str,'.JPG');
    im=imread(str);
    im=rgb2gray(im);
    im=imresize(im,[256 256]);
    temp=reshape(im,65536,1);
    z=[z temp];m=mean(z,2);
end
for i=1:size(z,2)
    t=double(z(i,:))-m;
    ta=[ta t];
end
 R=ta'*ta;    
 [v,d]=eig(R);   
  L_eig_vec=[];   
 for i=1:size(v,2)      
  if(d(i,i)>1)      
      L_eig_vec=[ L_eig_vec ;  v(:,i) ];   
     end   
 end    
 Eigenfaces = ta*L_eig_vec;
     disp('eig finished');
 projectedimages = []; 
Train_Number = size(ta,2);
 for i = 1 : Train_Number 
temp = Eigenfaces'*ta(:,i); %he
 projectedimages = [projectedimages temp];
end
target=eye(Train_Number);
save latest.mat
end

Good luck.

[Bereket Ayalew]

i use matlab 2013a

'' Error using -

Matrix dimensions must agree.

Error in Eigenfaces (line 14)

t=double(z(i,:))-m; ''.............. still says this

i also checked all the file extension and its correct

[madhan ravi]

Sulaymon Eshkabilov's answer moved here for consistency:

Hi,

The problem is resolved. I have picked up my old pc with MATLAB 2013b and managed to generate your problem. The problem was with the accuracy level and algorith used in eigen-value computation with eig(). In later versions of MATLAB, the most appropriate (accurate) algorithm is chosen automatically, but in older versions this is not the case. Also, a size selection of arrays has been changed in later version - it is rather automatic and well accustomated without a user notice. Anyhow, I have done some small changes and adjusted the code to MATLAB 2013b (2013a - hopefully) with eigenvalue computing algorithm with Choleky decomposition.

Here is the complete code:

function [  ] = Eigenfaces( ) 
z=[];
ta=[];
for ii=1:4
    str=int2str(ii);
    str=strcat(str,'.jpg');
    im=imread(str);
    im=rgb2gray(im);
    im=imresize(im,[256 256]);
    temp=reshape(im,65536,1);
    z=[z temp]; m=mean(z,2);
 end
for i=1:size(z,2)
    t=double(z(:,i))-m;
    ta=[ta t];
end
 R=ta'*ta;    
 [v,d]=eig(R, R, 'chol'); 
  L_eig_vec=[];   
   for i=1:size(v,2)      
  if(d(i,i)>1)      
      L_eig_vec=[ L_eig_vec ;  v(i,i) ];   
  end   
 end    
  Eigenfaces = ta*L_eig_vec;
     disp('eig finished');
 projectedimages = []; 
Train_Number = size(ta,2);
 for i = 1 : Train_Number 
temp = Eigenfaces'*ta(:,i); 
 projectedimages = [projectedimages temp];
end
target=eye(Train_Number);
save latest.mat
end

Good luck. Don't forget Thumbs UP ... :)

Answer 3

[Sulaymon Eshkabilov]

Hi,

The problem is resolved. I have picked up my old pc with MATLAB 2013b and managed to generate your problem. The problem was with the accuracy level and algorith used in eigen-value computation with eig(). In later versions of MATLAB, the most appropriate (accurate) algorithm is chosen automatically, but in older versions this is not the case. Also, a size selection of arrays has been changed in later version - it is rather automatic and well accustomated without a user notice. Anyhow, I have done some small changes and adjusted the code to MATLAB 2013b (2013a - hopefully) with eigenvalue computing algorithm with Choleky decomposition.

Here is the complete code:

function [  ] = Eigenfaces( ) 
z=[];
ta=[];
for ii=1:4
    str=int2str(ii);
    str=strcat(str,'.jpg');
    im=imread(str);
    im=rgb2gray(im);
    im=imresize(im,[256 256]);
    temp=reshape(im,65536,1);
    z=[z temp]; m=mean(z,2);
 end
for i=1:size(z,2)
    t=double(z(:,i))-m;
    ta=[ta t];
end
 R=ta'*ta;    
 [v,d]=eig(R, R, 'chol'); 
  L_eig_vec=[];   
   for i=1:size(v,2)      
  if(d(i,i)>1)      
      L_eig_vec=[ L_eig_vec ;  v(i,i) ];   
  end   
 end    
  Eigenfaces = ta*L_eig_vec;
     disp('eig finished');
 projectedimages = []; 
Train_Number = size(ta,2);
 for i = 1 : Train_Number 
temp = Eigenfaces'*ta(:,i); 
 projectedimages = [projectedimages temp];
end
target=eye(Train_Number);
save latest.mat
end

Good luck. Don't forget Thumbs UP ... :)