In dlib fhog_object_detector() represents a histogram-of-oriented-gradients based object detector. This object also implements a trained SVM classifier that is used for the final prediction. That is -HOG- for feature extraction and -SVM- for classification. I am using the default trained get_frontal_face_detector() to initialize a face detector and I want to extract the SVM parameters that implements.
I can dump and load the detector (both in .svm or .pkl) but I assume that the file has all the parameters inside (HOG parameters + SVM paramters).
import dlib
# initialize the detector (HOG + SVM)
detector = dlib.get_frontal_face_detector()
# Dump the detector
detector.save('detector.svm')
# Load back the detector
detector = dlib.fhog_object_detector('detector.svm')
How can I separate (or extract) the SVM parameters only using python or C++ ?? How can I Interpret the weights that represent the SVM's hyper-plane?
Thanks.
