save_known_face.py

# This python script can be used to save new face in face_encodings dir.
# If you want to save a new face encoding
#  python save_known_face.py -f image.jpg -n "ibrahim"
#                       OR
#  python save_known_face.py -f image.png -n "ibrahim"


import cv2
import numpy as np
import dlib
import argparse

def main(file, person_name, detect_faces, predictor):
    # Grab a single frame from WebCam
    frame = cv2.imread(file)
            
    # Find all the faces and face enqcodings in the frame
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    faces = detect_faces.detectMultiScale(gray, 1.3,5)
    face_locations = []
    for (x,y,w,h) in faces:
    	face_locations.append(dlib.rectangle(x, y, x+w, y+h))

    # If there is atleast 1 face then do face recognition 
    if (len(face_locations) > 0 and len(face_locations) < 2):
        for face_location in face_locations:
            face_encoding = predictor(frame, face_location)
            face_descriptor = face_rec.compute_face_descriptor(frame, face_encoding, 1)
            # Save encoding in a numpy file
            np.save("face_encodings/"+person_name, np.array(face_descriptor))
    else:
        print("Either no face or more then one face detected. Please check the image file again")
                
if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument("-f", "--file", type=str, help="give file for to extract face from it", required=True)
    parser.add_argument("-n", "--name", type=str, help="give name for the person", required=True)

    args = parser.parse_args()

    file = args.file
    person_name = args.name

    detect_faces = cv2.CascadeClassifier('haarcascade_frontalface_alt2.xml')
    predictor = dlib.shape_predictor("shape_predictor_68_face_landmarks.dat")
    face_rec = dlib.face_recognition_model_v1('dlib_face_recognition_resnet_model_v1.dat')

    main(file, person_name, detect_faces, predictor)