I am using ORL face Database for face recognition. There are 40 subjects each with 10 images each. The size of each image is 112 * 92 pixels. For feature extraction and dimensionality reduction, I used 2D- PCA. The feature matrix is now 112*8.
Following is the code after feature extraction --
% TrainFaces --- shape is 240 * 112 *8 (contains 240 images (8 images of 40 persons))
% ValFaces --- shape is 80 * 112 *8 (contains 80 images (2 images of 40 persons))
% TestFaces --- shape is 80 * 112 *8 (contains 80 images (2 images of 40 persons))
% train_target -- shape is 40 * 240 (1st bit is 1 if it is 1st image and all rest zero)
% val_target -- shape is 40 * 240
% test_target -- shape is 40 * 240
EigenFaces = [TrainFaces ; ValFaces ; TestFaces];
target = [train_target val_target test_target];
%%%%%%%%%% Now we can convert feature matrix into feature vector %%%%%%%%%%%%%
input = [];
for i = 1:size(EigenFaces,1)
- input = [input reshape(EigenFaces(i,:,:),size(EigenFaces,2)*size(EigenFaces,3),1)];
end
%%%%%%%% Creating Network and Training Them %%%%%%%%%%%
setdemorandstream(491218382);
net = patternnet(25);
net.divideFcn = 'divideind';
[net.divideParam.trainInd, net.divideParam.valInd, net.divideParam.testInd] = divideind(400,1:240,241:320,321:400);
net.trainFcn = 'trainrp';
[net,tr] = train(net,input,target);
%%%%%%%%%%%%% Testing the network %%%%%%%%%%%%%%%
predict = 0; % Number of correctly identified images out of 80
for i =1:80
- [tar_val tar_ind] = max(test_target);
- [out_val out_ind] = max(net(reshape(TestFaces(i,:,:),size(TestFaces,2)*size(TestFaces,3),1)));
- if tar_ind == out_ind
- predict = predict +1;
- end
end
predict
The predict always shows zero. Kindly help me analyse the graphs, thanks in advance.
Edit : I have uploaded the EigenFaces and Target values in the link given below. Kindly have a look.
http://goo.gl/DjfTWY
Ouelli Abdelhaq : Thank you for your help. But what variables are you talking about? Do you mean code?
I have attached the Eigenfaces and target values in the link shared above and also here. Kindly have a look.
http://goo.gl/DjfTWY
It's simple. Construct the confusion matrix (en.wikipedia.org/wiki/Confusion_matrix) for the classifier output. Measure the parameters Precision, Recall, F-Measure for each class and the balanced accuracy.
I hope you know how to calculate these. If not, please let me know.
Hi, I agree with Mr Kumar, the confusion function measures the precision parameters, here is the code:
[c,cm] = confusion(testtarget,Netoutput);
fprintf('Percentage Correct Classification : %f%%\n', 100*(1-c));
fprintf('Percentage Incorrect Classification : %f%%\n', 100*c);
fprintf('Number of correctly identified images out of 80 :%i \n', sum(diag(cm)));
else, i have some notes:
1- the number of features is too much (896);
2-iI think you must use an other technique for features extraction and then use the PCA for feature number reduction.
3-finally, your database is too small.
then you can try the code in attachement, the maximun number of correctly identified images was 23 but you may improve that.
Thank you Abdelhaq and Kumar for your insight and help. I would definitely work as suggested.
@ Abdelhaq :
1) Can you suggest any suitable feature extraction technique ? I know there are lot. Actually I am working on implementing face recognition algorithm (neural network) on FPGA chip. The modelling in Matlab is only for gaining information about neural network topology. So I couldn't do a detailed literature review.
2) I think you are right. As the number of features are huge, the network may not be learning properly. Can you give an rough idea about the number of features that should be used as input for the neural network to train properly?
Zernike descriptor perform well for imagae feature extraction, but certaily there is some algorithmes specified in face recognition, image processing is not my field.
the number of features is huge,i suggest after feature extraction you perform PCA to reduce the number to 60 or 40.
- Badges
- Science topic