machine learning - HOG Feature Extraction of Arabic Line Images -

i doing project on writer identification. want extract hog features line images of arabic handwriting. , use gaussian mixture model classification. link database containing line images : http://khatt.ideas2serve.net/

so questions follows;

1. there 3 folders namely test, train , validate. so, folder need extract features. , purpose should use each of folders.
2. do need extract features individual images , merge them or there method extract features of images together.

test, train , validate

read stats se question: what difference between test set , validation set?

this basic machine learning, should go , review course literature, since seems you're missing pretty important machine learning concepts.

do need extract features individual images , merge them or there method extract features of images together.

it seems, again, you're missing basic concepts here. histogram of oriented gradients subdivides image , finds oriented gradient. see question examples of hos looks.

the traditional way of using hog is: each image in training set, extract hog, use these train svm, validate training validation set, use trained svm on test set.