LBPH prediction slow?

asked 2017-06-23 02:25:44 -0600

Skap95 gravatar image

Hi, I made a face recognition program with the LBPH Face Recognizer and it works pretty well, despite this i saw that the more train images you got, the prediction became slower and slower... Is there any solution to make the prediction faster?
I work on a Raspberry Pi 3 and the LBPH parameters that i currently use are 1,8,8,8,130.0 and I currently got about 600 pictures for the training

Thank you in advance!

PS: sorry for my English, i'm French !

edit retag flag offensive close merge delete

Comments

how large are your images ? also, what is your use-case ? (is an identification / closest of many scheme really , what you need ?)

berak gravatar imageberak ( 2017-06-23 02:27:45 -0600 )edit
1

@berak My images are 200*200, and i use the face recognition to identifie a person in front of a webcam video, but i need to be able to recognize a lot of people (currently i only have 20 person in the database)

Skap95 gravatar imageSkap95 ( 2017-06-23 02:34:08 -0600 )edit

sorry, misread it. image size is more or less irrelevant here.

berak gravatar imageberak ( 2017-06-23 02:48:35 -0600 )edit
1

there is no really optimisation in predict method for LBPH. When you have N training faces predict calulates N times compareHist

LBerger gravatar imageLBerger ( 2017-06-23 04:17:57 -0600 )edit
add a comment