close all

rootFolder = 'att_faces';

image_dims = [112, 92];

imageFolders = dir(rootFolder);

imageFolders = {imageFolders.name};

imageFolders = imageFolders(startsWith(imageFolders(:), 's'));

images = [];

subjectIds = {};

for i = 1:size(imageFolders, 2)

folder = imageFolders{i};

subjectPath = strcat(rootFolder, '/',folder);

images = dir(subjectPath);

images = {images.name};

images = images(or(endsWith(images(:), 'pgm'), endsWith(images(:), 'png')));

subjectId = folder;

subjectIds = unique([subjectIds subjectId]);

faces(i).id = subjectId;

idNumber = subjectId(2:end);

faces(i).idNumber = str2num(idNumber);

faces(i).pictures = {};

faces(i).vectors = [];

for j = 1:size(images, 2)

fullImagePath = strcat(subjectPath, '/', images{j});

image = imread(fullImagePath);

image = double(image);

imageIndex = images{j};

imageIndex = imageIndex(1:end-4);

imageIndex = str2num(imageIndex);

faces(i).pictures{imageIndex} = image;

% read and convert images to column vector

currentPicture = faces(i).pictures{imageIndex};

[irow icol] = size(currentPicture);

faceVector = reshape(currentPicture',irow*icol,1);

faces(i).vectors = [faces(i).vectors faceVector];

end

% Reconstructed average face

faceFields = fieldnames(faces);

faceCell = struct2cell(faces);

faceSize = size(faceCell);

faceCell = reshape(faceCell, faceSize(1), []); % Px(MxN)

faceCell = faceCell'; % (MxN)xP

faceCell = sortrows(faceCell, 2);

faceCell = reshape(faceCell', faceSize);

faces = cell2struct(faceCell, faceFields, 1);

train_data = [];

test_data = [];

test_data_id = [];

train_data_id = [];

test_index = [1];

for i = 1:size(faces,2)

for j = 1:size(faces(i).vectors, 2)

faceVector = faces(i).vectors(:,j);

if ismember(j, test_index)

test_data = [test_data, faceVector];

test_data_id = [test_data_id getNumberId(faces(i).id)];

else

train_data = [train_data, faceVector];

train_data_id = [train_data_id getNumberId(faces(i).id)];

end

end

mean_face = mean(train_data, 2);

avgFace = reshape(mean_face, 92, 112);

figure; imshow(uint8(avgFace'));

Class_number = (size(faces, 2)); % Number of classes (or persons)

Class_population = size(faces(1).vectors,2) - size(test_index,2); % Number of images in each class

P = Class_population * Class_number; % Total number of training images

shifted_images = train_data - repmat(mean_face, 1, P);

run('face_recognition_training.m')

run('face_recognition_test.m')

run('face_recognition_evaluation.m')