I wanted to do image recognition with tensorflow, so I thought about making a program that cuts out only the face.
I ran the program by referring to this site (https://newtechnologylifestyle.net/opencv_face_trimingu/).
On this site, it was a program that cuts only the face for one image, so if you cut all the images in the file, the program will be
I modified it, but it is an error and I do not understand the problem.
I'm sorry, but I would like to ask for your help.

Source code

import cv2
import glob
directory ='D: /program/face/*.jpg' // File with face image
path ='D:/program/facecut /'// Output destination file
HAAR_FILE ='face.xml'
cascade = cv2.CascadeClassifier (HAAR_FILE)
c = 0
for image_picture in glob.glob ('./face/*.jpg'):
    img = cv2.imread (image_picture)
    img_g = cv2.imread (image_picture, 0)
    face = cascade.detectMultiScale (img_g)
    print (face)
    for x, y, w, h in face:
        face_cut = img [y: y + h, x: x + w]
    c + = 1
    filename = path +'face_cut {} .jpg'.format (c)
    cv2.imwrite (filename, face_cut)

Traceback (most recent call last):
  File "d: /program/OneImageFaceCut.py", line 19, in<module>
    cv2.imwrite (filename, face_cut)
NameError: name'face_cut' is not defined

What I tried

for image_picture in glob.glob ('./face/*.jpg'):
    img = cv2.imread (image_picture)
    img_g = cv2.imread (image_picture, 0)
    face = cascade.detectMultiScale (img_g)
    print (face)
    for x, y, w, h in face:
        face_cut = img [y: y + h, x: x + w]
    c + = 1
filename = path +'face_cut {} .jpg'.format (c)
cv2.imwrite (filename, face_cut)

In this way, I put out cv2.imwrite () outside the for statement. Then, the compilation passed, and only one image with the face cut out was saved.
I can imagine that only one sheet is output, but I'm not sure if it won't compile when I put it in the for statement.

Supplementary information (FW/tool version, etc.)

opencv 4.0.1