I am using skimage learn in python to extract HOG features from an image.
skimage.features.hog returns a tuple (fd,hog_arr) where fd stands for HOG feature descriptors and 'hog_arr' is for visual representation of HOG features. (This is what I understood after looking at http://scikit-image.org/docs/dev/auto_examples/plot_hog.html)
I want to apply HOG on the entire image and extract the specific local HOG features out of it. For example, if the image is a face, I want to apply HOG on face and extract the hog features corresponding to 'eye' of the facial image.
As fd variable stores the actual HOG features of the image, I am not understanding how to extract such specific HOG features corresponding to a specific location inside the image ( I am able to visualize the HOG features corresponding to the image through 'hog_arr' through cv2.imshow)
How to go about this? Any help?