[OpenCV-Python]35.OpenCV face detection and recognition - Example 2 (face recognition training local data)

35. Face detection and recognition of OpenCV - Example 2 (face recognition training local data)


  face detection refers to the process of locating faces in images. Face recognition is to further judge people's identity on the basis of face detection. This example is a summary of the previous codes related to face detection and recognition.

1, Collect face samples

  put the face image to be recognized in the face folder:

  photos placed in each folder are as follows:

import os
import cv2
import imghdr
import numpy as np
from imutils import *

# Face detection and saving
def facedetect(image, output):
    # Get file name
    name = os.path.basename(image)
    # Read in picture
    image = cv2.imread(image)
    # Change to grayscale
    image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
    # Cascade classifier, face detection
    detector = cv2.CascadeClassifier("haarcascade_frontalface_alt.xml")
    rects = detector.detectMultiScale(image, scaleFactor=1.1, minNeighbors=3, minSize=(20, 20), flags=cv2.CASCADE_SCALE_IMAGE)
    # Loop each face
    for (x,y,w,h) in rects:
        # The face is intercepted and transformed into a fixed size of 200 * 200
        f = cv2.resize(image[y:y+h, x:x+w], (200,200))
        # Write to specified path
        cv2.imwrite(os.path.join(output, name), f)

# Detect and intercept face
def predict_face(path, output):
    # If the folder does not exist, create a folder
    if not os.path.exists(output):
    # Cycle the pictures under each character's folder
    for files in os.listdir(path):
        # Detect whether it is a folder
        if os.path.isdir(os.path.join(path, files)):
            # Define the output path of the intercepted face
            output2 = os.path.join(output, files)
            # If the folder does not exist, create a folder
            if not os.path.exists(output2):
            # Full path to people folder
            files = os.path.join(path, files)
            # Cycle each photo of each person
            for file in os.listdir(files):
                # Photo full path
                file = os.path.join(files, file)
                # Detect face and save
                facedetect(file, output2)

# Sample path and acquisition path
predict_face('faces', 'predict_faces')

  get predict after running the program_ Faces folder, which contains the detection pictures of predicted faces. After opening, you need to delete the pictures with recognition errors (the red box is the recognition error image):

2, Generate Label

  the program will be based on predict_ According to the number of people in the faces subfolder, labels starting from 0 are automatically generated.

# Generate label file
def get_label(path):
    fh = open("label.txt", 'w')
    # Represent face label
    label = 0
    for root, dirs, files in os.walk(path):
        # Cycle through each folder
        for subdir in dirs:
            # Folder full path
            subdir_path = os.path.join(root,subdir)
            # Cycle through each photo below each people folder
            for file in os.listdir(subdir_path):
                # Photo full path
                filepath = os.path.join(subdir_path, file)
                # Determine whether the file type is a picture type
                imgType = imghdr.what(filepath)
                if imgType == 'jpeg' or imgType == 'png':
                    # Save picture path
                    # label
            # Everyone's label is different. Count from 0
            label = label + 1            
# Path to collect face Tags

  generated label Txt will generate corresponding labels according to the classified paths in the folder, as shown below:

3, Train your own data model

   when calling the face recognizer, the program will read in the known image and the corresponding label corresponding to the label file for training, and save the training file at the end.

# Save picture data
images = []
# Save label
labels = []
# Open file
fh = open("label.txt")
# Loop each line
for line in fh:
    # To; Split string
    arr = line.split(";")
    # Part 0 is the picture path to read the file
    img = cv2.imread(arr[0],0)
    # Save picture data
    # Save the corresponding label data
# model = cv2.face.EigenFaceRecognizer_create()
# model = cv2.face.FisherFaceRecognizer_create()
model = cv2.face.LBPHFaceRecognizer_create()

# Training model
model.train(np.array(images), np.array(labels))
# Save model

  generated label file and saved model:

4, Face recognition (picture file)

   prepare unknown face images to be predicted in advance and put them into the test folder.

   after detecting the face, pass the model Predict () method.

# Define character names
name= ['KobeBryant','LeBronJames','TracyMcGrady']

# Load the trained model

# Read in the test picture to do the test
for file in os.listdir('test'):
    file = os.path.join('test', file)
    # Determine file type
    imgType = imghdr.what(file)
    if imgType == 'jpeg' or imgType == 'png'or 'jpg':
        # Read in picture
        image = cv2.imread(file)
        # Change to grayscale
        gray = cv2.cvtColor(image,cv2.COLOR_BGR2GRAY)
        # cascade classifier
        detector = cv2.CascadeClassifier("haarcascade_frontalface_alt.xml")
        rects = detector.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=3, minSize=(20, 20), flags=cv2.CASCADE_SCALE_IMAGE)
        # Loop each face
        for (x,y,w,h) in rects:
            # Draw a rectangular box
            cv2.rectangle(image, (x,y), (x+w,y+h), (255,255,255), 2)
            # Face recognition
            face = cv2.resize(gray[y:y+h,x:x+w], (200,200))
            # Predict people
            params = model.predict(face)
            # Write the character's name
        i += 1

  the recognition results are as follows:

5, Face recognition using video

  the model generated by the picture file can be recognized by the camera.

import cv2

capture = cv2.VideoCapture(0)     

frame_width = capture.get(cv2.CAP_PROP_FRAME_WIDTH)
frame_height = capture.get(cv2.CAP_PROP_FRAME_HEIGHT)
fps = capture.get(cv2.CAP_PROP_FPS)

if capture.isOpened() is False:  
    print('CAMERA ERROR !')
model = cv2.face.LBPHFFaceRecognizer_create()

name= ['Kobe Bryant','LeBron James','Tracy McGrady']


detector = cv2.CascadeClassifier("haarcascade_frontalface_alt.xml")   

while capture.isOpened():
    ret, frame = capture.read()   
    if ret is True:              
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)     
        rects = detector.detectMultiScale(gray, scaleFactor=1.1, minNeighbors=3, minSize=(20, 20), flags=cv2.CASCADE_SCALE_IMAGE)
        for (x,y,w,h) in rects:

            cv2.rectangle(frame, (x,y), (x+w,y+h), (255,255,255), 2)

            face = cv2.resize(gray[y:y+h,x:x+w], (200,200))

            params = model.predict(face)

        cv2.imshow('FACE', frame)  
        k = cv2.waitKey(100)
        if k == ord('q'):


  the recognition results are as follows:

6, Opencv Python resource download

Opencv Python test pictures, Chinese official documents, opencv-4.5.4 source code


   the above content introduces the code example summary of OpenCV Python face detection and recognition. Articles on python, data science and artificial intelligence will be published from time to time. Please pay more attention and connect three times with one button (● '◡' ●).

Tags: Python OpenCV Computer Vision opencv-python

Posted by walkonet on Thu, 06 Jan 2022 06:03:57 +1030