Open aaferrero opened 3 years ago
The cost will depend on the size of your images, since the detection part will dominate the timing. A few suggestions: Try one of the 3 available face detector (see option face_detector= ), resize images where appropriate, make sure you init fa object only once.
I´ve read a lot about smaller models in other issues. @1adrianb i know you get asked a lot, but are the smaller models coming?
Greetings
fa = face_alignment.FaceAlignment(face_alignment.LandmarksType._2D, flip_input=False, device='cuda') fa.get_landmarks(img)
how much seconds is it cost? I get 8 seconds for 3 pictures.