How to give Image as Input to SVM

Let image is (n*n) matrix. If i change it to (N^2*1) matrix,

the SVMtrain gives error " SVM must have 2 column Matrix". To change the Image into two columns doesnt look logical. I want Face recognition using MultuSVM but unable to get the idea of storing image in format. (X1, X2) with Classes(yi).

[Ahmed]

Please post the line that calls the 'svmtrain' function, the sizes of your data and label matrix/vector and the exact error message. The line "SVM must have 2 column Matrix" does not sound like literally taken from the command prompt output.

[Pratik Oak]

You have to give the features of an image as input to svmtrain. Not direct image.

Answer 1

[Pratik Oak]

Hi we cant give image as input to svm. U should calculate feature/features of an image. Give that features matrix as input to svm.

[Ahmed]

Under certain circumstances it is possible to input the images directly. For instance, if the images are registered. However, for face recognition this probably won't work (sufficiently) well.