Turn from https://blog.csdn.net/chen801090/article/details/95072592 , the original text is to identify Huang Jiaju and Huang Jiaqiang. The difference is that this article changes to Wanxi and Jiang Shuying. Why do you choose them... In addition, add a recognizer to download in the link (the resource has not been approved). Nothing else is different, just put it in your blog. In addition, I read the comments and said that the algorithm in the link above is not recognized correctly. Maybe because there are too few training sets, I also met with the pictures of two goddesses. Then I changed the picture

# # -*- coding:utf-8 -*-
import cv2
import os
import numpy as np

# Create a mapping list of tags and people names (tags can only be integers)
subjects = ["jiangshuying", "wanxi"]

# Face detection
def detect_face(img):
    # Convert the test image to grayscale image, because opencv face detector needs grayscale image
    gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

    # Load OpenCV face detection classifier Haar
    face_cascade = cv2.CascadeClassifier('/home/menglingwei/Desktop/study_1/haarcascade_frontalface_default.xml')

    # To detect multi-scale image, the return value is a list of facial region information (x,y, width, height)
    faces = face_cascade.detectMultiScale(gray, scaleFactor=1.2, minNeighbors=5)

    # If no face is detected, the original image is returned
    if (len(faces) == 0):
        return None, None

    # At present, it is assumed that there is only one face, xy is the coordinate of the upper left corner, wh is the width and height of the rectangle
    (x, y, w, h) = faces[0]

    # Back to the front of the image
    return gray[y:y + w, x:x + h], faces[0]


# This function will read all training images, detect faces from each image and return two lists of the same size, namely face information and label
def prepare_training_data(data_folder_path):
    # Get directories in the data folder (one for each topic)
    dirs = os.listdir(data_folder_path)

    # Two lists save all faces and labels respectively
    faces = []
    labels = []

    # Browse each directory and access the images in it
    for dir_name in dirs:
        # Dir? Name (STR type) is the label
        label = int(dir_name)
        # Create a directory path that contains images of the current theme
        subject_dir_path = data_folder_path + "/" + dir_name
        # Get the image name in the given topic directory
        subject_images_names = os.listdir(subject_dir_path)

        # Browse each picture and detect the face, then add the face information to the face list faces []
        for image_name in subject_images_names:
            # Establish image path
            image_path = subject_dir_path + "/" + image_name
            # Read image
            image = cv2.imread(image_path)
            # Display image 0.1s
            cv2.imshow("Training on image...", image)
            cv2.waitKey(100)

            # Face detection
            face, rect = detect_face(image)
            # We ignore undetected faces
            if face is not None:
                # Add a face to the face list and label it accordingly
                faces.append(face)
                labels.append(label)

    cv2.waitKey(1)
    cv2.destroyAllWindows()#Close window after training
    # The final return value is face and tag list
    return faces, labels


# Draw a rectangle on the image according to the given (x, y) coordinates and width height
def draw_rectangle(img, rect):
    (x, y, w, h) = rect
    cv2.rectangle(img, (x, y), (x + w, y + h), (128, 128, 0), 2)


# Identify the person name according to the given (x, y) coordinate
def draw_text(img, text, x, y):
    cv2.putText(img, text, (x, y), cv2.FONT_HERSHEY_COMPLEX, 1, (128, 128, 0), 2)


# This function identifies the person in the transferred image and draws a rectangle and its name around the detected face
def predict(test_img):
    # Make a copy of the image so you can keep the original image
    img = test_img.copy()
    # Face detection
    face, rect = detect_face(img)
    # Predictive face
    label = face_recognizer.predict(face)
    # Get the name of the corresponding label returned by the face recognizer
    label_text = subjects[label[0]]

    # Draw a rectangle around the detected face
    draw_rectangle(img, rect)
    # Name the forecast
    draw_text(img, label_text, rect[0], rect[1] - 5)
    # Return the predicted image
    return img


# Call the prepare? Training? Data() function to train the model
faces, labels = prepare_training_data("training_data")

# Create LBPH recognizer and start training, or choose Eigen or Fisher recognizer
face_recognizer = cv2.face.LBPHFaceRecognizer_create()
face_recognizer.train(faces, np.array(labels))


# Load test image
test_img1 = cv2.imread("test_data/test1.jpg")
test_img2 = cv2.imread("test_data/test2.jpg")

# Execution prediction
predicted_img1 = predict(test_img1)
predicted_img2 = predict(test_img2)

# Show two images
cv2.imshow(subjects[0], predicted_img1)
cv2.imshow(subjects[1], predicted_img2)
cv2.waitKey(0)
cv2.destroyAllWindows()

It can be recognized normally. Post scholars should pay attention to it