import io
import picamera
import cv2
import numpy
import serial

#Load a cascade file for detecting faces
face_cascade = cv2.CascadeClassifier('/home/pi/opencv-3.3.0/data/haarcascades/haarcascade_frontalface_default.xml')

ser = serial.Serial('/dev/ttyACM0')

while True:
    #Create a memory stream so photos doesn't need to be saved in a file
    stream = io.BytesIO()

    #Get the picture (low resolution, so it should be quite fast)
    #Here you can also specify other parameters (e.g.:rotate the image)
    with picamera.PiCamera() as camera:
        camera.resolution = (320, 240)
        camera.capture(stream, format='jpeg')

    #Convert the picture into a numpy array
    buff = numpy.fromstring(stream.getvalue(), dtype=numpy.uint8)

    #Now creates an OpenCV image
    image = cv2.imdecode(buff, 1)

    #Convert to grayscale
    gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)

    #Look for faces in the image using the loaded cascade file
    faces = face_cascade.detectMultiScale(gray, 1.1, 5)
    print("Found "+str(len(faces))+" face(s)")

    if len(faces) > 0:
        ser.write(b'f')
    else:
        ser.write(b'n')