Does locating the face first then encoding the face have higher accuracy than just encoding the face?

[Ann5t]

Since I noticed that there's this attribute number_of_times_to_upsample in the function face_locations, I wonder if locating the face first have a slower outcome but a higher accuracy?

e.g. locating the face first then encoding the face

face_locations = face_recognition.face_locations(image, 
                                                 number_of_times_to_upsample=2, 
                                                 model='hog')
face_encodings_in_image = face_recognition.face_encodings(image, 
                                                          known_face_locations=face_locations,
                                                          num_jitters=1, 
                                                          model='small')

e.g. just encoding the face

face_encodings_in_image = face_recognition.face_encodings(image, 
                                                          known_face_locations=face_locations,
                                                          num_jitters=1, 
                                                          model='small')