SVM classifier on digits dataset

Hi,

I have to use SVM classifier on digits dataset. The dataset consists of images of digits 28x28 and a toal of 2000 images.

I tried to use svmtrain but the matlab gave an error that svmtrain has been removed. so now i am using fitcsvm.

My code is as below:

labelData = zeros(2000,1);

for i=1:1000

labelData(i,1)=1;

end

for j=1001:2000

labelData(j,1)=1;

end

SVMStruct =fitcsvm(trainingData,labelData)

%where training data is the set of images of digits.

I need to know how i can predict the outputs of test data using svm? Further is my code correct?

Answer 1

[Ridwan Alam]

fitcsvm() is not designed to train directly on 28x28 images. You need to extract feature vectors for each image. After feature extraction, you will have a dataset with size 2000xN, where N is the number of features for each image. Let's say X is this new extracted feature set of size 2000xN, and Y is your labeledData of size 2000x1.

Now, before training fitcsvm(), you need to use cvpartition() (or something similar, maybe your custom method) to divide both the features and the labels data into training and testing dataset. Then, you can use

SVMStruct =fitcsvm(trainingData,trainingLabelData) 

For more details: https://www.mathworks.com/help/stats/support-vector-machine-classification.html

As far as I know, you can use Convolutional Neural Networks (CNN) to train directly on images.

Minor issues with your code:

1. in both loops, you are assigning the labelData the same value.
2. you don't really need those loops: labelData(1:1000) = 0; labelData(1001:2000) = 1;

Hope this helps.