OpenCV face detection is slow on Raspberry Pi

Question

I am testing Raspberry Pi with OpenCV and Python coding. The video streaming works great (medium speed), but when I run face detection on the stream the CPU is pegged and refreshing the image is slow.

Here is what I have. How can I optimize my code?

#!/usr/bin/env python
import sys
import cv2.cv as cv
from optparse import OptionParser
min_size = (20, 20)
image_scale = 2
haar_scale = 1.2
min_neighbors = 2
haar_flags = 0

def detect_and_draw(img, cascade):
    # allocate temporary images
    gray = cv.CreateImage((img.width,img.height), 8, 1)
    small_img = cv.CreateImage((cv.Round(img.width / image_scale),
                               cv.Round (img.height / image_scale)), 8, 1)
                               cv.Round (img.height / image_scale)), 8, 1)

    # convert color input image to grayscale
    cv.CvtColor(img, gray, cv.CV_BGR2GRAY)

    # scale input image for faster processing
    cv.Resize(gray, small_img, cv.CV_INTER_LINEAR)

    cv.EqualizeHist(small_img, small_img)

    if(cascade):
        t = cv.GetTickCount()
        faces = cv.HaarDetectObjects(small_img, cascade, cv.CreateMemStorage(0),
                                     haar_scale, min_neighbors, haar_flags, min_size)
        t = cv.GetTickCount() - t
        print "detection time = %gms" % (t/(cv.GetTickFrequency()*1000.))
        if faces:
            for ((x, y, w, h), n) in faces:
                # the input to cv.HaarDetectObjects was resized, so scale the 
                # bounding box of each face and convert it to two CvPoints
                pt1 = (int(x * image_scale), int(y * image_scale))
                # bounding box of each face and convert it to two CvPoints
                pt1 = (int(x * image_scale), int(y * image_scale))
                pt2 = (int((x + w) * image_scale), int((y + h) * image_scale))
                cv.Rectangle(img, pt1, pt2, cv.RGB(255, 0, 0), 3, 8, 0)

    cv.ShowImage("result", img)

if __name__ == '__main__':

    parser = OptionParser(usage = "usage: %prog [options] [camera_index]")
    parser.add_option("-c", "--cascade", action="store", dest="cascade", type="str", help="Haar cascade file, default %default", default = "/usr/local/share/OpenCV/haarcascades")
    (options, args) = parser.parse_args()

    cascade = cv.Load(options.cascade)
    capture = cv.CreateCameraCapture(0)
    cv.NamedWindow("result", 1)

    if capture:
        frame_copy = None
        while True:
            frame = cv.QueryFrame(capture)
            if not frame:
                cv.WaitKey(0)
                break
            if not frame_copy:
                frame_copy = cv.CreateImage((frame.width,frame.height),
                                            cv.IPL_DEPTH_8U, frame.nChannels)
            if frame.origin == cv.IPL_ORIGIN_TL:
                cv.Copy(frame, frame_copy)
            else:
                cv.Copy(frame, frame_copy)
            else:
                cv.Flip(frame, frame_copy, 0)

            detect_and_draw(frame_copy, cascade)

            if cv.WaitKey(10) != -1:
                break
    else:
        image = cv.LoadImage(input_name, 1)
        detect_and_draw(image, cascade)
        cv.WaitKey(0)

    cv.DestroyWindow("result")

Answer 1

I can suggest you the LBP cascade instead of Haar. It is known to be up to 6 times faster with very close detection rate.

But I'm not sure if it is accessible in legacy python interface. The cv2.CascadeClassifier class from new wrappers can do detection for LBP cascade.