TestTogetherForever.m

%% Together Forever
% TestTogetherForever in order to mix both (TestHoGG & TestMoodPCA), this will recognize you
%       from a new snap and identify your "mood" from that one.


% Uncomment to do this test apart
% clc
% clear
% close all
% name = []

%% WebCam setup
% This is commented because it was chosen at GUI_Menu
%     1 para la webCam normal
%     2 para la webCam de mi tablet
%             CamType=1;

%% Name Verification
name;
if isempty(name) == 1
    name = inputdlg('Introduzca su nombre por favor: ','Bienvenido');
end


%% See if its necessary to do new snaps

Snap_folder =0;

% Init mood labels
mood6 = {'Happy', 'Bored', 'Crazy', 'Angry', 'Sleepy', 'Sad'};
mood9 = {'Happy', 'Bored', 'Crazy', 'Angry', 'Sleepy', 'Sad', 'Smart' , 'Thankful' , 'Weird'};

 answer1 = BoxMenu ('Menu', 'Do you want new snaps?');
            
        if answer1 == 1  
            choice = questdlg('Number of New Photos?', ...
        'New Snap Pool','6','9', 'fix');
        % Handle response
            switch choice
                case '6'
                    disp([choice ' snaps chosen!'])
                    num = 6;
                case '9'
                    disp([choice ' snaps chosen'])
                    num = 9;
            end
            rmdir('Together4Ever/BagOfSnaps/You','s');
            mkdir('Together4Ever/BagOfSnaps/You');
            Snap_folder = 'Together4Ever/BagOfSnaps/You';
            
            if num == 6
                Img.mood = mood6;
            elseif num == 9
                Img.mood = mood9;
            end
            
            
            for i=1:num
               uiwait(msgbox({'Let`s' Img.mood{i} 'snap'}, 'Be Smart plz','warn','modal'));
               [~, IFaces, bboxes] = Snapshot (CamType);
               I = HOGFeatures(IFaces,bboxes);
               I=ReSize(I);
               imwrite(I,['Foto_', num2str(i),'.jpg']);
               movefile(strcat('Foto_', num2str(i),'.jpg'),Snap_folder);
                
            end
        end
        
        
        if answer1 == 2
            rmdir('Together4Ever/BagOfSnaps/You','s');
            mkdir('Together4Ever/BagOfSnaps/You');
            % Select a folder with 6 or 9 photos to train them
            uiwait(msgbox('Choose "Together4Ever/Training" folder PLZ', 'Be Smart plz','warn','modal'));
            disp('Selección de Repositorio de tus fotos, sugerencia "Together4Ever/Training"');
            folder_name = uigetdir('Together4Ever/Training','selecciona el repositorio');
            Snap_folder = 'Together4Ever/BagOfSnaps/You';
            
        end
        
        
%  Se sale del programa        
        if answer1 == 0
           GUI_Menu
           return; 
        end
        
        close all
 
%% Create the imageSet and copying 

 if answer1 == 2
        MeDatabase = imageSet(folder_name,'recursive');
        % Convertimos todas las fotos al mismo tamaño y creamos una copia en el
        % directorio que se ha elegido
        for i=1:MeDatabase.Count
            % Lectura de la imagen
            I = read(MeDatabase(1),i);
            I = ReSize (I);
            imwrite(I,['Foto_', num2str(i),'.jpg']);
            movefile(strcat('Foto_', num2str(i),'.jpg'),Snap_folder);
        end
 end

%% Show a montage of the training photos
FaceDatabase = imageSet(Snap_folder, 'recursive');
figure;
montage(FaceDatabase.ImageLocation);
title('Images of YOU');

%% Sacamos las HOGFeatures para entrenarlas
TrainingDatabase = imageSet('Together4Ever/BagOfSnaps','recursive');
featureCount =1;
for i=1:size(TrainingDatabase,2) % Para contar los 41 Folders
     for j = 1:TrainingDatabase(i).Count
        trainingFeatures(featureCount,:) = extractHOGFeatures(read(TrainingDatabase(i),j));
        trainingLabel{featureCount} = TrainingDatabase(i).Description;    
        featureCount = featureCount + 1;
    end
    personIndex{i} = TrainingDatabase(i).Description;
end

%% Creating class using fitcecoc
faceClassifier = fitcecoc(trainingFeatures,trainingLabel);

%% Start MoodPCA Part
dim= size(read(FaceDatabase(1),1),1); % 224  ; % image size
M=FaceDatabase.Count; %9 o 6;        % number faces
            if M == 6
                Img.mood = mood6;
            elseif M == 9
                Img.mood = mood9;
            end
for i=1:M
    
    Img.data{i} = read(FaceDatabase(1),i); %Creating an structure with pixel data

end
save classFile Img;

load classFile;
% ScreenSize to plot
scrsz = get(0,'ScreenSize');    
ancho = scrsz(3);               
alto = scrsz (4);
ancho = round (ancho/3);
alto = round (alto/2)-40;

num=FaceDatabase.Count;
dim=size(read(FaceDatabase(1),1),1);

%% Create z vector with data info    
if num==9
    z= [ Img.data{1}  Img.data{2}    Img.data{3}; Img.data{4}     Img.data{5}  Img.data{6};...
        Img.data{7} Img.data{8} Img.data{9}];
    figure(1),imshow(z,'Initialmagnification','fit');title('Faces')
end

if num==6
    z= [ Img.data{1}  Img.data{2}    Img.data{3}; Img.data{4}     Img.data{5}  Img.data{6}];
    figure(1),imshow(z,'Initialmagnification','fit');title('Faces')
end

%% Compute Mean
averageImg=zeros(dim);
for i=1:num
    Img.data{i} = im2single(Img.data{i});
    averageImg   = averageImg  + (1/num)*Img.data{i};
end
% Uncommen to show the figure
%figure(2),imshow(averageImg,'Initialmagnification','fit');title('average')

%% Normalize (removing mean)
for i=1:num
    Img.dataAvg{i}  = Img.data{i} -averageImg;
end

if num == 9
    z  = [ Img.dataAvg{1}  Img.dataAvg{2}   Img.dataAvg{3}  ; Img.dataAvg{4}     Img.dataAvg{5}  Img.dataAvg{6};...
        Img.dataAvg{7}     Img.dataAvg{8}  Img.dataAvg{9}];
    % Uncommen to show the figure
%     figure(3),imshow(z,'Initialmagnification','fit');title('z average')
end

if num == 6
    z  = [ Img.dataAvg{1}  Img.dataAvg{2}   Img.dataAvg{3}  ; Img.dataAvg{4}     Img.dataAvg{5}  Img.dataAvg{6}];
    % Uncommen to show the figure
%     figure(3),imshow(z,'Initialmagnification','fit');title('z average')
end

%% Calc Eigenvectors
% Cenerate A = [ img1(:)  img2(:) ...  imgM(:) ]; calculo de los autovectores
A = zeros(dim*dim,num);% (N*N)*num   2500*4
for i=1:num
    A(:,i) = Img.dataAvg{i}(:);
end
% Covariance matrix small dimension (transposed)
Cov_mat = A'*A;
% Uncommen to show the figure
% figure(4),imagesc(Cov_mat);title('covariance')

%% Eigenvectros  in small dimension
[ Eigvec_V, Eigval_D ]  = eig(Cov_mat);% v num*num e num*num only diagonal 4 eigen values
% eigan face in large dimension  A*Eigvec_V is eigen vector of Clarge
Large_V = A*Eigvec_V;% 2500*num*num*num  =2500 *num
% reshape to eigen face
Eigenfaces=[];
for i=1:num
    c_arr  = Large_V(:,i);
    Eigenfaces{i} = reshape(c_arr,dim,dim);
end
diag_eigval=diag(Eigval_D);
[xc,xci]=sort(diag_eigval,'descend');% largest eigenval

if num == 6
    z  = [ Eigenfaces{xci(1)}  Eigenfaces{xci(2)}   Eigenfaces{xci(3)} ; Eigenfaces{xci(4)}     Eigenfaces{xci(5)}   Eigenfaces{xci(6)}];
    % Uncommen to show the figure
%     figure(5),imshow(z,'Initialmagnification','fit');title('Eigenfaces')
end

if num == 9
    z  = [ Eigenfaces{xci(1)}  Eigenfaces{xci(2)}   Eigenfaces{xci(3)} ; Eigenfaces{xci(4)}     Eigenfaces{xci(5)}   Eigenfaces{xci(6)};...
       Eigenfaces{xci(7)}  Eigenfaces{xci(8)}   Eigenfaces{xci(9)} ];
   % Uncommen to show the figure 
%    figure(5),imshow(z,'Initialmagnification','fit');title('Eigenfaces')
end
%% Weights
nsel=M% select  eigen faces
for img_num=1:num  % image number
  for i=1:nsel   % eigen face for coeff number
    img_weight(img_num,i) =   sum(A(:,img_num).* Eigenfaces{xci(i)}(:)) ;
  end
end



% Para el final
%% Classify new IMG


NotFinished = false;

while ~NotFinished 
    close (GUI_TogetherForever)
    answer2 = BoxMenu ('Let´s trying to find Yourself','Let''s take a new picture to compare?');
    
    switch answer2
            case 1
              [I, IFaces, bboxes] = Snapshot (CamType);
              % Get HOGFeatures of the photo  and cropp
              [Icrop, hogFeature, visualization] = HOGFeatures (IFaces, bboxes);
              testFace = Icrop;
            case 2
               uiwait(msgbox('Then, choose one to compare :)', 'He he he','warn','modal'));
               [file, location] = uigetfile ('*jpg') ;
               I = imread(strcat(location,file));
               % Image proccesing
               faceDetector = vision.CascadeObjectDetector;
               bboxes = step(faceDetector, I);
               IFaces = insertObjectAnnotation(I, 'rectangle', bboxes, 'Face');
               % Get HOGFeatures of the photo  and cropp
               [Icrop, hogFeature, visualization] = HOGFeatures(IFaces,bboxes);
               testFace = Icrop;
            case 0
              close all
              uiwait(msgbox('Bye byeeeeee', 'Haleluyaaa','warn','modal')); 
              close (GUI_TogetherForever)
              GUI_Menu
              return;  
                
    end 
     
    
    %% EigenFaces 
    testFace = imresize(testFace,[dim dim]);
    testFace   =  im2single(testFace);
%     Uncommen to show the image
%     figure(6), imshow(testFace,'Initialmagnification','fit'); title('Test Face')
    Average_face = testFace(:)-averageImg(:); % normilized face
    for(i=1:nsel)
      face_weight(i)  =  sum(Average_face.* Eigenfaces{xci(i)}(:)) ;
    end

    % compute distance
    for img_num=1:num  
        face_sumcur=0;
        for(i=1:nsel)
            face_sumcur = face_sumcur + (face_weight(i) -img_weight(img_num,i)).^2;
        end
        diffWeights(img_num) =   sqrt( face_sumcur);
    end
%     Commen to show figures
    close all    
    
    % Difference of EigenFaces
    [val in]=min(diffWeights);
    diffWeights  = diffWeights.';
    [min2, pos] = sort(diffWeights(:));
    val2 = min2(2);
    in2 = pos(2);
    diff2 = val2 - val;
   
    
    %% Probamos si funciona el clasificador
    personLabel = predict (faceClassifier, hogFeature);
    value = char(personLabel);
    
%% Load EmojiDatabase
    EmojiDatabase = imageSet('Emojis','recursive');
	for i=1:EmojiDatabase.Count
        Img.emoji{i} = read(EmojiDatabase(1),i);
    end
	
 %% Print Results 
    %  Uncommen to show results
    %     switch value
    %         case 'You'
    %             % Eigen Results
    %              if diff2 < 100
    %                 figure('Position', [2*ancho 10 ancho alto]);  
    %                 imshow([Img.data{in},Img.data{in2}]);
    %                 title(['Your mood is...' Img.mood{in} '   and   ' Img.mood{in2}]);
    %              else   
    %                 figure('Position', [2*ancho 10 ancho alto]), imshow(Img.data{in}),title(['Your mood is...' Img.mood{in}]);
    %              end
    %             %HOGG Results
    %             h= msgbox(personLabel,'And the result is...');
    %             set(h, 'position', [100 440 400 100]); %makes box bigger
    %             ah = get( h, 'CurrentAxes' );
    %             ch = get( ah, 'Children' );
    %             set( ch, 'FontSize', 14 ); %makes text bigger
    %     
    %         case 'NotYou'
    %             % Eigen Results
    %              if diff2 < 100
    %     %              Modificar
    %                 figure('Position', [2*ancho 10 ancho alto]);  
    %                 imshow([Img.emoji{in},Img.emoji{in2}]);
    %                 title(['Your mood is...' Img.mood{in} '   and   ' Img.mood{in2}]);
    %              else   
    %                 figure('Position', [2*ancho 10 ancho alto]), imshow(Img.emoji{in}),title(['Your mood is...' Img.mood{in}]);
    %              end
    %             %HOGG Results
    %             h= msgbox(personLabel,'And the result is...');
    %             set(h, 'position', [100 440 400 100]); %makes box bigger
    %             ah = get( h, 'CurrentAxes' );
    %             ch = get( ah, 'Children' );
    %             set( ch, 'FontSize', 14 ); %makes text bigger
    %         
    %     end

% Show results with GUI
        run GUI_TogetherForever.m
        pause (10)


end