the difference between HOG Detector and HOG Descriptor

See more:

MatLab

OpenCV

image-processing

I am trying to implement the Hog Descriptor using java without using Hog descriptor which is implemented in opencv, but I have the following questions:

Are HOG Detector and HOG Descriptor the same?

After we extract interest points or key points from an image from what do we extract HOG descriptor? Form the whole image or from the interest points / key points which we detected before from image.

When we compute the gradient do we divide the image into blocks before or after we compute the image gradient?

Regards Thanks in advance