What should the image look like after applying the HOG algorithm?

Question

Algorithms

Good day, I don’t really understand how the b / w image should look after applying the HOG algorithm HOG
is a method of directional gradients
Gradient is a vector
A vector is a segment that has a direction, simply speaking an arrow
But in all implementations of HOG in python I I saw only the cv2.line method (opencv library), that is, a line was drawn, and I don’t understand how in this case to determine in which direction the pixels darken, and why a line is drawn instead of an arrow

Answer 1

This algorithm extracts descriptors that can be used for ML. The original image itself remains untouched. Arrows are needed only for clarity of the algorithm.
Translation of an article with visual illustrations

spoiler