main_desktop.py

from imutils import face_utils #for resizing
import numpy as np
import argparse
import imutils
import dlib
import cv2
import time
from scipy.spatial import distance as dist #euclidian distance
import pandas as pd
import csv
from pathlib import Path


csv_columns = ['name', 'face_data']
csv_file = 'all_face_data.csv'


camera_id = 0
EAR_AR_THRESH = 0.3
CONSEC_FRAMES = 1
TOTAL = 0
COUNT = 0
(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"]
(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]

print('App is preparing, please wait.')
detector = dlib.get_frontal_face_detector() # detect the faces in the image. How many faces are there
predictor = dlib.shape_predictor('./shape_predictor_68_face_landmarks.dat') # predict the face landmarks such as mouth or eyes
#(lStart, lEnd) = face_utils.FACIAL_LANDMARKS_IDXS["left_eye"] 
#(rStart, rEnd) = face_utils.FACIAL_LANDMARKS_IDXS["right_eye"]
facerec = dlib.face_recognition_model_v1('./dlib_face_recognition_resnet_model_v1.dat') #pretrained model. 
#we send the data to this function and it returns a 128D vector that described the faces.



def eye_aspect_ratio(eye):
    #https://www.pyimagesearch.com/wp-content/uploads/2017/04/blink_detection_plot.jpg
    a = dist.euclidean(eye[1], eye[5])
    b = dist.euclidean(eye[2], eye[4])
    
    c = dist.euclidean(eye[0], eye[3])
    
    ear = (a+b) / (2.0 * c)
    
    return ear


def write_dict_to_csv(csv_file, csv_columns, dict_data):
    try:
        with open(csv_file, 'w') as csvfile:
            writer = csv.DictWriter(csvfile, fieldnames=csv_columns)
            #writer.writeheader()
            for key, value in dict_data.items():
                writer.writerow({'name': key, 'face_data': value})
    except IOError:
        print("I/O error", csv_file)
    return


def append_to_csv(csvfile, data):
    with open(csvfile, 'a') as f:
        writer = csv.writer(f)
        for key, value in data.items():
            writer.writerow([key,value])
    return


def cvt_to_array(data, split_with=''):
    if split_with == '':
        return np.array(list(map(float, data)))
    else:
        return np.array(list(map(float, data.split(split_with))))


def menu():
    print('Welcome to face recognition system. What do you want to do?\n')
    action = 0
    while action != 3:
        print('1) Run the app\n')
        print('2) Save new face to database\n')
        print('3) Exit\n')
        action = int(input('>>'))
        if action == 1:
            run_app()
        elif action == 2:
            save_face()
    


def run_app():
    global COUNT, TOTAL
    found_face = 0
    cap = cv2.VideoCapture(camera_id)
    while True:
        ret, image = cap.read()
        image = imutils.resize(image, width=300)
        gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
        if found_face % 5 == 0:
            rects = detector(gray, 0)
        for (i, rect) in enumerate(rects):
            # determine the facial landmarks for the face region, then
            # convert the landmark (x, y)-coordinates to a NumPy array
            shape = predictor(gray, rect)
            
            trying = np.array(facerec.compute_face_descriptor(image, shape))
            #distance_faces = dist.euclidean(face_data, trying)
            my_file = Path("./" + csv_file)
            with open(my_file, 'r') as f:
                reader = csv.reader(f)
                for row in reader:
                    if row == [] or row[1] == "face_data":
                        continue
                    else:
                        #row[1] = np.array(list(map(float, row[1].split('\n'))))
                        row[1] = cvt_to_array(row[1], '\n')
                        trying = cvt_to_array(trying)
                        distance_faces = dist.euclidean(row[1], trying)
                        if distance_faces < 0.55:
                            content = row[0]
                            break
                        else:
                            content = "unknown"
                        
            cv2.putText(image,content, (10,40), cv2.FONT_HERSHEY_PLAIN, 1, (0,255,0), 2)
            
            
            shape = face_utils.shape_to_np(shape)
            for (x, y) in shape:
                cv2.circle(image, (x, y), 1, (0, 0, 255), -1)
           
            leftEye = shape[lStart:lEnd]
            rightEye = shape[rStart:rEnd]
            leftEAR = eye_aspect_ratio(leftEye)
            rightEAR = eye_aspect_ratio(rightEye)
            leftEyeHull = cv2.convexHull(leftEye)
            rightEyeHull = cv2.convexHull(rightEye)
            cv2.drawContours(image, [leftEyeHull], -1, (0, 255, 0), 1)
            cv2.drawContours(image, [rightEyeHull], -1, (0, 255, 0), 1)
            ear = (leftEAR + rightEAR) / 2.0
            
            
            if ear < EAR_AR_THRESH:
                COUNT += 1
            else:
                if COUNT >= CONSEC_FRAMES:
                    TOTAL += 1
                
                COUNT = 0
            cv2.putText(image, "Blinks: {}".format(TOTAL), (300, 40),cv2.FONT_HERSHEY_SIMPLEX, 1, (0, 0, 255), 2)
            
        cv2.imshow("Frame", image)
        key = cv2.waitKey(1) & 0xFF
        #break
        # if the `q` key was pressed, break from the loop
        if key == ord("q"):
            break
        #time.sleep(0.1)
    cap.release()
    cv2.destroyAllWindows()

def save_face():
    

    #capture the person and save as the 128D vector
    # this part captures only once
    cap = cv2.VideoCapture(camera_id)
    #while True:

    face_data = []
    labels = []
    data = {}


    face_number = 0
    while face_number == 0:
        print('Please show your whole face to camera. When the face is detected, you will be asked for the name.')
        time.sleep(0.5)
        ret, image = cap.read()
        image = imutils.resize(image, width=500) #resizing
        gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) #it should convert to gray in onder to improve resultt.
        rects = detector(gray, 0) # detect how many faces in the image
        
        for (i, rect) in enumerate(rects): 
            # for every faces
            # determine the facial landmarks for the face region, then
            # convert the landmark (x, y)-coordinates to a NumPy array
            shape = predictor(gray, rect) # predict the face landmarks in image. 

            face_descriptor = facerec.compute_face_descriptor(image, shape) # send the shape data to resnet model. it returns a 128D vector
            
            while face_descriptor == -1:
                print('Face not found.')
            else:
                face_data.append(face_descriptor) # save the face data to array
                shape = face_utils.shape_to_np(shape)
                for (x, y) in shape:
                    cv2.circle(image, (x, y), 1, (0, 0, 255), -1)
                name = str(input('Who is this : '))
                while (not name.isalpha or name == ''):
                    print('Please enter a valid name.')
                    name = str(input('Who is this : '))
                labels.append(name)
                data[labels[0]] = face_data[0]
                face_data=[]
                labels=[]
                my_file = Path("./" + csv_file)
                if my_file.is_file():
                    append_to_csv(csv_file, data)
                    print('File already exist, data is appended to file')
                else:
                    write_dict_to_csv(csv_file, csv_columns, data)    
                    print('File has been created and data saved to file.')
                face_number += 1
            #print(face_descriptor)
            '''

            leftEye = shape[lStart:lEnd]
            rightEye = shape[rStart:rEnd]
            leftEAR = eye_aspect_ratio(leftEye)
            rightEAR = eye_aspect_ratio(rightEye)
            leftEyeHull = cv2.convexHull(leftEye)
            rightEyeHull = cv2.convexHull(rightEye)
            cv2.drawContours(image, [leftEyeHull], -1, (0, 255, 0), 1)
            cv2.drawContours(image, [rightEyeHull], -1, (0, 255, 0), 1)
            ear = (leftEAR + rightEAR) / 2.0



            ''' 
    cv2.imshow("Saved face", image)
    cv2.waitKey(0)
    #key = cv2.waitKey(1) & 0xFF
    #break
    # if the `q` key was pressed, break from the loop
    #if key == ord("q"):
    #    break
    #time.sleep(0.5)
    cap.release()
    cv2.destroyAllWindows()


if __name__ == '__main__':
    menu()