camera_face_scroll.py

'''This python code performs face tracking and simulates scrolling. First, the image feed is recieved from the camera. A facial detection algorithm is applied
once in every 5 frames. On detecting the face, a rectangular frame is drawn around the detected face, and the deviation from the center of the screen to the center of the face is taken as the scroll delta.
Requires Python + OpenCV'''
import cv2
from pynput.mouse import Button, Controller

mouse = Controller()


HAAR_CASCADE_PATH = "haarcascade_frontalface_alt.xml"
CAMERA_INDEX = 0


def detect_faces(image):
    # print('detect fn' #)
    faces = []
    # detected = cv.HaarDetectObjects(image, cascade, storage, 1.2, 2,
    # cv.CV_HAAR_DO_CANNY_PRUNING, (100,100))
    detected = cascade.detectMultiScale(
        image, 1.3, 4, cv2.CASCADE_SCALE_IMAGE, (20, 20))
    # print(detected)

    if detected != []:
        # print('face detected' #)
        for (x, y, w, h) in detected:  # for (x,y,w,h),n in detected:
            faces.append((x, y, w, h))
    return faces


def get_motion(face):
    # yaw is x-axis - horizontal axis
    # pitch is y-axis - depth axis
    # roll is z-axis - vertical axis

    #[0][0] - x, [0][1] - y, [0][2] - w, [0][3] - h

    threshold = 40

    # w,h are approx constant for U,D,L,R events
    # checking if w,h in range of origin(w,h)+/-5
    if (face[0][2] > (origin[0][2] - threshold)) and (face[0][2] < (origin[0][2] + threshold)) and (face[0][3] > (origin[0][3] - threshold)) and (face[0][3] < (origin[0][3] + threshold)):

        # check x while y is same
        if face[0][1] > (origin[0][1] - threshold) and face[0][1] < (origin[0][1] + threshold):
            if face[0][0] > (origin[0][0] - threshold) and face[0][0] < (origin[0][0] + threshold):
                # user is in origin location
                print('origin')
                return 25  # no motion
            else:
                if (face[0][0] - origin[0][0]) > 0:
                    # LEFT motion event - S button
                    print('LEFT')
                    return 9
                elif (face[0][0] - origin[0][0]) < 0:
                    # RIGHT motion event - A button
                    print('RIGHT')
                    return 8
        else:
            # check y while x is same
            if (face[0][1] - origin[0][1]) > 0:
                # DOWN motion event - Q button
                print("DOWN")
                mouse.scroll(0, -20)

                return 6
            elif (face[0][1] - origin[0][1]) < 0:
                # UP motion event - W button
                print('UP')
                mouse.scroll(0, 20)

                return 7
    else:
        # possible events: Zoom in, Zoom out
        if (face[0][2] - origin[0][2]) > 0:
            # ZOOM IN motion event - = button
            mouse.scroll(0, -20)
            return 4
        elif (face[0][2] - origin[0][2]) < 0:
            # ZOOM OUT motion event - -button
            mouse.scroll(0, 20)
            return 5


if __name__ == "__main__":
    # print('creating window' #)
    cv2.namedWindow("Video")

    capture = cv2.VideoCapture(CAMERA_INDEX)
    # storage = cv.CreateMemStorage()
    cascade = cv2.CascadeClassifier(
        cv2.data.haarcascades + "haarcascade_frontalface_default.xml")

    i = 0
    c = -1

    global ctr, origin, faces

    faces = []
    ctr = 0
    origin = []

    while (c == -1):

        retval, image = capture.read()
        # print('acq img frm' #)

        # Only run the Detection algorithm every 5 frames to improve
        # performance
        if i % 5 == 0:
            # print('calling detect' #)
            ctr += 1
            faces = detect_faces(image)

            if(faces != [] and origin != []):
                dir = get_motion(faces)
                print(dir)

        if ctr == 20:
            # approx 3 secs of config time
            origin = faces
            print('origin is ', origin)

        for (x, y, w, h) in faces:
            # print('drawing rectangle' #)
            cv2.rectangle(image,
                          (x, y), (x + w, y + h), 255)

        # print('showing img' #)
        cv2.imshow("Video", image)
        # cv.ShowImage("Video", image)
        i += 1
        c = cv2.waitKey(10)
        if(c == 27):
            # escape
            break

haarcascades can be found in opencv repo, just google.