Facial_Recognition_Inference.py

import os, sys, time
import dlib
import cv2
import numpy as np
import DBHelper
import Upload_Thief


def inference():
    try:
        import cPickle  # Python 2
    except ImportError:
        import _pickle as cPickle  # Python 3

    pwd = sys.path[0]
    PREDICTOR_PATH = pwd + '/Facial_models/shape_predictor_68_face_landmarks.dat'
    FACE_RECOGNITION_MODEL_PATH = pwd + '/Facial_models/dlib_face_recognition_resnet_model_v1.dat'

    SKIP_FRAMES = 1
    THRESHOLD = 0.4

    faceDetector = dlib.get_frontal_face_detector()
    shapePredictor = dlib.shape_predictor(PREDICTOR_PATH)
    faceRecognizer = dlib.face_recognition_model_v1(FACE_RECOGNITION_MODEL_PATH)

    index = np.load(pwd + '/Facial_models/index.pkl', allow_pickle=True)
    faceDescriptorsEnrolled = np.load(pwd + '/Facial_models/descriptors.npy')

    cam = cv2.VideoCapture(0)
    count = 0

    x1 = x2 = y1 = y2 = 0

    cond = False
    thief = False
    label = 'unknown'
    start = time.time()

    while DBHelper.get_power() == "on":
        success, im = cam.read()

        if not success:
            print('cannot capture input from camera')
            break

        if (count % SKIP_FRAMES) == 0:

            img = cv2.cvtColor(im, cv2.COLOR_BGR2RGB)
            faces = faceDetector(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))
            minDistance = 0.0

            for face in faces:

                shape = shapePredictor(cv2.cvtColor(img, cv2.COLOR_BGR2RGB), face)

                x1 = face.left()
                y1 = face.top()
                x2 = face.right()
                y2 = face.bottom()

                faceDescriptor = faceRecognizer.compute_face_descriptor(img, shape)

                # dlib format to list
                faceDescriptorList = [m for m in faceDescriptor]
                # to numpy array
                faceDescriptorNdarray = np.asarray(faceDescriptorList, dtype=np.float64)
                faceDescriptorNdarray = faceDescriptorNdarray[np.newaxis, :]

                # Euclidean distances
                distances = np.linalg.norm(faceDescriptorsEnrolled - faceDescriptorNdarray, axis=1)

                # Calculate minimum distance and index of face
                argmin = np.argmin(distances)  # index
                minDistance = distances[argmin]  # minimum distance

                if minDistance <= THRESHOLD:
                    label = DBHelper.get_firstname(index[argmin]) + "_" + DBHelper.get_lastname(index[argmin])
                    cond = True
                else:
                    label = 'unknown'
                    cond = False

                # print("time taken = {:.3f} seconds".format(time.time() - t))

            cv2.rectangle(im, (x1, y1), (x2, y2), (0, 255, 0), 2)
            font_face = cv2.FONT_HERSHEY_SIMPLEX
            font_scale = 0.8
            text_color = (0, 255, 0)
            printLabel = '{} {:0.4f}'.format(label, minDistance)
            cv2.putText(im, printLabel, (int(x1), int(y1)), font_face, font_scale, text_color, thickness=2)

        cv2.imshow('img', im)

        k = cv2.waitKey(1) & 0xff
        if k == 27:
            break

        count += 1
        if cond:
            DBHelper.set_motor("on")
            DBHelper.set_alarm("off")
            start = time.time()
        elif not cond:
            DBHelper.set_motor("off")
            DBHelper.set_alarm("on")
            end = time.time()
            if end - start >= 30:
                thief = True
                DBHelper.set_power("off")

    DBHelper.set_alarm("off")
    DBHelper.set_motor("off")
    cv2.destroyAllWindows()
    if thief:
        Upload_Thief.upload_thief_face()


if __name__ == "__main__":
    inference()