How to feed image to svm classifier?

Question

Support vector machine

There is an image - a person's face, it is translated into a hog descriptor and has a size of 128 × 256, that is, in general, the image has 32768 pixels with a value of either 0 (black) or 255 (white), the hog descriptor looks something like this: a training sample of 3000 faces in the hog descriptor and 3000 non-faces (benches, streets, buildings, etc.) also in the hog descriptor
Please explain to me the following things:
I understand how to train a classifier on such a sample
The implementation is supposed to be in python, sklearn.svm and other library implementations are not interested, the task is to do it yourself I will

add my question: I
read about 20 more articles, the first specific question appeared:
How to represent an image as a point in n-dimensional space, that is, how to transform the input image into one of these points?