MATLAB Answers

0

Transfer Learning Using AlexNet

Asked by Peter Mayhew on 27 Sep 2017
Latest activity Answered by Mark Sherstan on 17 Dec 2018
I have followed the Transfer Learning Using AlexNet tutorial and sucessfully classified a new collection of images. https://uk.mathworks.com/help/nnet/examples/transfer-learning-using-alexnet.html
I would now like to apply the transfer learning so my webcam can recognise the new images, as shown in this tutorial. https://uk.mathworks.com/matlabcentral/fileexchange/60659-deep-learning-in-11-lines-of-matlab-code
How can I combine my classification of images with AlexNet?
Peter

  0 Comments

Sign in to comment.

Products

1 Answer

Answer by Mark Sherstan on 17 Dec 2018
 Accepted Answer

A really simple example is as follows:
vid = videoinput('winvideo', 1);
set(vid, 'ReturnedColorSpace', 'RGB');
img = getsnapshot(vid);
labelMachine(img,net,classifier) % net and classifier are from your transfer learning on alexnet
function label = labelMachine(img,net,classifier)
featureLayer = 'fc7';
% Pre-process the image as required for the CNN
img = imresize(img, [227 227]);
% Extract image features using the CNN
imageFeatures = activations(net, img, featureLayer);
% Make a prediction using the classifier
label = char(predict(classifier, imageFeatures));
end
I did a project that used a webcam and alex net transfer learning which can be found here if you need further referances.

  0 Comments

Sign in to comment.