algorithm.py

import face_recognition
import numpy as np
from person import Person
from PIL import Image
import cv2

THRESHOLD_NORMALIZER = float(42.5)
THRESHOLD_WEIGHT = float(1.0)

MIN_IMAGE_HEIGHT = 10
MIN_IMAGE_WIDTH = 10

def isImageValid(height, width):
	return ((height >= MIN_IMAGE_HEIGHT) and (width >= MIN_IMAGE_WIDTH))

def read_image_from_disk(name):
		return face_recognition.load_image_file("data/" + name)

def getSumIndex(encoding):
    return np.sqrt(np.sum(encoding**2))

def threshold(val):
    val = THRESHOLD_WEIGHT + (val/THRESHOLD_NORMALIZER)
    return val
	# return 1

def getEncoding(image, sharpness, image_size):
	encoding = face_recognition.face_encodings(image,
					known_face_locations=[(0,image_size[0], image_size[1],0)])
	# encoding = face_recognition.face_encodings(image)
	# print "encoding: ", encoding
	if len(encoding) == 0:
		return None

	encoding = encoding[0]
	encoding = np.array(encoding)
	
	thresh = threshold(sharpness)
	return encoding*thresh

def getPilImage(image):
	array = cv2.cvtColor(np.array(image), cv2.COLOR_RGB2BGR)
	pil_image = Image.fromarray(array)
	return pil_image

def get_sharpness(pil_image):
	pil_image = pil_image.convert('L')
	array = np.asarray(pil_image, dtype=np.int32)
	
	gy, gx = np.gradient(array)
	gnorm = np.sqrt(gx**2 + gy**2)
	sharpness = np.average(gnorm)
	return float(sharpness)

def getFaces(image, timestamp):
	faces = face_recognition.face_locations(image)
	person_list = list()
	for location in faces:
		top, right, bottom, left = location
		# print "Face Location {} {} {} {}".format(top,right,bottom,left)
		face_image = image[top:bottom, left:right]
		pil_image = getPilImage(face_image)
		width, height = pil_image.size

		if not isImageValid(height, width):
			continue

		'''Calculating Params'''
		face_sharpness = get_sharpness(pil_image)
		# TODO: Pass the known face locations for better accuracy
		face_encoding = getEncoding(face_image, face_sharpness, [width, height])

		if face_encoding is None:
			# print "Face Encoding Failed"
			continue

		face_sum_index = getSumIndex(face_encoding)

		person = Person(sharpness=face_sharpness, sum_index=face_sum_index,
						encodings=face_encoding, pil_image=pil_image, timestamp=timestamp, 
						img_array=face_image)
		person_list.append(person)
	return person_list

def compare(encoding, matching_list_of_encodings):
	return face_recognition.compare_faces(matching_list_of_encodings, encoding)