-
Notifications
You must be signed in to change notification settings - Fork 0
/
face_trainer.py
33 lines (29 loc) · 1.47 KB
/
face_trainer.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
import cv2
import numpy as np
from PIL import Image
import os
# Path for face image database
path = 'Path to folder where face were saved'
recognizer = cv2.face.LBPHFaceRecognizer_create() #LBPH (LOCAL BINARY PATTERNS HISTOGRAMS) Face Recognizer is included in the OpenCV package.
detector = cv2.CascadeClassifier("C:\\Python\\Lib\\site-packages\\cv2\\data\\haarcascade_frontalface_default.xml")
# function to get the images and label data
def get_images_and_labels(paths):
image_paths = [os.path.join(paths, f) for f in os.listdir(paths)] # To go through each image
face_samples = []
ids = []
for image_path in image_paths:
pil_img = Image.open(image_path).convert('L') # convert it to grayscale
img_numpy = np.array(pil_img, 'uint8')
id2 = int(os.path.split(image_path)[-1].split(".")[0]) # Getting images of a user to be appended together
faces = detector.detectMultiScale(img_numpy)
for (x, y, w, h) in faces:
face_samples.append(img_numpy[y:y+h, x:x+w])
ids.append(id2)
return face_samples, ids
print("\n [INFO] Training faces. It will take a few seconds. Wait ...")
faces1, id1 = get_images_and_labels(path)
recognizer.train(faces1, np.array(id1))
# Save the model into trainer/trainer.yml
recognizer.write('C:\\docx\\trainer\\trainer.yml')
# Print the number of faces trained and end program
print("\n [INFO] {0} faces trained. Exiting Program".format(len(np.unique(id1))))