Finished A3 Q1 roughly, might revisit

generate_cropped_notfaces.m

@@ -32,7 +32,7 @@
 
     % Crop image
     croppedImage = img(startRow:startRow+dim-1, ...
-                        startCol:startCol+dim-1);
+                        startCol:startCol+dim-1, :);
 
     % Save image
     imgPath = strcat(new_imageDir, '/', int2str(n_have), '.jpg');
@@ -42,4 +42,45 @@
     % list of uncropped images
     n_have = n_have + 1;
     i = i + 1;
-end
+end
+
+%% Split training images into a training set and validation set
+percentTraining = 80;
+numOfTraining = floor((percentTraining/100)*n_have);
+
+% Making the directories
+facesDir = 'cropped_training_images_faces';
+facesList = dir(sprintf('%s/*.jpg',facesDir));
+
+notFacesDir = 'cropped_training_images_notfaces';
+notFacesList = dir(sprintf('%s/*.jpg',notFacesDir));
+
+[~,~,~] = mkdir(strcat(facesDir,'/validation'));
+[~,~,~] = mkdir(strcat(facesDir,'/training'));
+[~,~,~] = mkdir(strcat(notFacesDir,'/validation'));
+[~,~,~] = mkdir(strcat(notFacesDir,'/training'));
+
+for i = 1:numOfTraining
+    % Copy faces training images
+    source = strcat(facesDir,'/',facesList(i).name);
+    dest = strcat(facesDir,'/training/',facesList(i).name);
+    copyfile(source, dest);
+
+    % Copy not faces training images
+    source = strcat(notFacesDir,'/',notFacesList(i).name);
+    dest = strcat(notFacesDir,'/training/',notFacesList(i).name);
+    copyfile(source, dest);
+end 
+
+for i = numOfTraining+1:n_have
+    % Copy faces validation images
+    source = strcat(facesDir,'/',facesList(i).name);
+    dest = strcat(facesDir,'/validation/',facesList(i).name);
+    copyfile(source, dest);
+
+    % Copy not faces validation images
+    source = strcat(notFacesDir,'/',notFacesList(i).name);
+    dest = strcat(notFacesDir,'/validation/',notFacesList(i).name);
+    copyfile(source, dest);
+end
+

get_features.m

@@ -1,6 +1,9 @@
-close all
-clear
-run('../vlfeat-0.9.20/toolbox/vl_setup')
+%% Run if VLFeat is not set up
+run('vlfeat-0.9.21/toolbox/vl_setup');
+
+%% Get Features
+close all;
+clear;
 
 pos_imageDir = 'cropped_training_images_faces';
 pos_imageList = dir(sprintf('%s/*.jpg',pos_imageDir));
@@ -10,15 +13,18 @@
 neg_imageList = dir(sprintf('%s/*.jpg',neg_imageDir));
 neg_nImages = length(neg_imageList);
 
-cellSize = 6;
-featSize = 31*cellSize^2;
+% cellSize needs to be a factor of 36
+cellSize = 4;
+im = imread(sprintf('%s/%s',pos_imageDir,pos_imageList(1).name));
+[imRows, imCols, ~] = size(im);
+featSize = 31*(imRows/cellSize)*(imCols/cellSize);
 
 pos_feats = zeros(pos_nImages,featSize);
 for i=1:pos_nImages
     im = im2single(imread(sprintf('%s/%s',pos_imageDir,pos_imageList(i).name)));
     feat = vl_hog(im,cellSize);
     pos_feats(i,:) = feat(:);
-    fprintf('got feat for pos image %d/%d\n',i,pos_nImages);
+%    fprintf('got feat for pos image %d/%d\n',i,pos_nImages);
 %     imhog = vl_hog('render', feat);
 %     subplot(1,2,1);
 %     imshow(im);
@@ -32,7 +38,7 @@
     im = im2single(imread(sprintf('%s/%s',neg_imageDir,neg_imageList(i).name)));
     feat = vl_hog(im,cellSize);
     neg_feats(i,:) = feat(:);
-    fprintf('got feat for neg image %d/%d\n',i,neg_nImages);
+%    fprintf('got feat for neg image %d/%d\n',i,neg_nImages);
 %     imhog = vl_hog('render', feat);
 %     subplot(1,2,1);
 %     imshow(im);
@@ -41,4 +47,4 @@
 %     pause;
 end
 
-save('pos_neg_feats.mat','pos_feats','neg_feats','pos_nImages','neg_nImages')
+save('pos_neg_feats.mat','pos_feats','neg_feats','pos_nImages','neg_nImages');

train_svm.m

@@ -1,13 +1,33 @@
-run('../vlfeat-0.9.20/toolbox/vl_setup')
-load('pos_neg_feats.mat')
+%% Run if VLFeat is not set up
+run('vlfeat-0.9.20/toolbox/vl_setup');
 
-feats = cat(1,pos_feats,neg_feats);
-labels = cat(1,ones(pos_nImages,1),-1*ones(neg_nImages,1));
+%% Split into training and validation
+close all;
+clear;
 
-lambda = 0.1;
-[w,b] = vl_svmtrain(feats',labels',lambda);
+load('pos_neg_feats.mat');
+
+percentTraining = 80;
+numOfTraining = floor((percentTraining/100)*pos_nImages);
+
+trainFeats = cat(1,pos_feats(1:numOfTraining,:), ...
+                    neg_feats(1:numOfTraining,:));
+validFeats = cat(1,pos_feats(1+numOfTraining:end,:), ...
+                    neg_feats(1+numOfTraining:end,:));
+trainLabels = cat(1,ones(numOfTraining,1), ...
+                    -1*ones(numOfTraining,1));
+validLabels = cat(1,ones(pos_nImages-numOfTraining,1), ...
+                    -1*ones(pos_nImages-numOfTraining,1));
+
+%% Train
+
+lambda = 0.06;
+[w,b] = vl_svmtrain(trainFeats',trainLabels',lambda);
 
 fprintf('Classifier performance on train data:\n')
-confidences = [pos_feats; neg_feats]*w + b;
+confidences = [pos_feats(1:numOfTraining,:); neg_feats(1:numOfTraining,:)]*w + b;
+[tp_rate, fp_rate, tn_rate, fn_rate] =  report_accuracy(confidences, trainLabels);
 
-[tp_rate, fp_rate, tn_rate, fn_rate] =  report_accuracy(confidences, labels);
+fprintf('Classifier performance on validation data:\n')
+confidences = [pos_feats(1+numOfTraining:end,:); neg_feats(1+numOfTraining:end,:)]*w + b;
+[tp_rate, fp_rate, tn_rate, fn_rate] =  report_accuracy(confidences, validLabels);

vlfeat-0.9.21/.gitattributes

@@ -0,0 +1,3 @@
+*.manifest -crlf -diff -merge
+*.vcproj binary
+*.sln binary

vlfeat-0.9.21/.gitignore

@@ -0,0 +1,57 @@
+# Mac OS X Finder
+.DS_Store
+
+# MEX files
+*.mexglx
+*.mexmac
+*.mexmaci
+*.mexmaci64
+*.mexa64
+*.mex
+
+# Python
+*.pyc
+
+# Other generated files and directories
+VERSION
+docsrc/version.html
+bin
+results
+doc
+build
+toolbox/mexw32
+toolbox/mexmaci
+toolbox/mexmaci64
+toolbox/mexmac
+toolbox/mexglx
+toolbox/mexa64
+toolbox/mexw64
+toolbox/mex
+toolbox/noprefix
+
+# Xcode
+vlfeat.xcodeproj/*.mode1
+vlfeat.xcodeproj/*.mode1v3
+vlfeat.xcodeproj/*.mode2v3
+vlfeat.xcodeproj/*.pbxuser
+vlfeat.xcodeproj/*.perspective*
+vlfeat.xcodeproj/*.xcworkspace
+vlfeat.xcodeproj/xcuserdata
+
+# Visual C++
+vlfeat.suo
+vlfeat.ncb
+vlfeat.vcproj.*.user
+vc/*
+
+# Emacs backups
+*~
+
+# Vim swap files
+.*.swp
+
+# Opt directory
+opt/*
+
+# iDraw files
+*.idraw

vlfeat-0.9.21/COPYING

@@ -0,0 +1,25 @@
+Copyright (C) 2007-11, Andrea Vedaldi and Brian Fulkerson
+Copyright (C) 2012-13, The VLFeat Team
+All rights reserved.
+
+Redistribution and use in source and binary forms, with or without
+modification, are permitted provided that the following conditions are
+met:
+1. Redistributions of source code must retain the above copyright
+   notice, this list of conditions and the following disclaimer.
+2. Redistributions in binary form must reproduce the above copyright
+   notice, this list of conditions and the following disclaimer in the
+   documentation and/or other materials provided with the
+   distribution.
+
+THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+"AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+(INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.