先安装依赖库dlib、face_recognition、cv2

下载wheel文件：

python3.6：

dlib-19.7.0-cp36-cp36m-win_amd64.whl: https://drfs.ctcontents.com/file/1445568/768652503/68cb5d/Python/dlib-19.7.0-cp36-cp36m-win_amd64.whl

python3.7:

dlib-19.17.99-cp37-cp37m-win_amd64.whl: https://drfs.ctcontents.com/file/1445568/768652504/b726a5/Python/dlib-19.17.99-cp37-cp37m-win_amd64.whl

python3.8:

dlib-19.19.0-cp38-cp38-win_amd64.whl.whl: https://drfs.ctcontents.com/file/1445568/768652508/77e657/Python/dlib-19.19.0-cp38-cp38-win_amd64.whl.whl

再使用pip安装face_recognition、cv2

pip install opencv-python
pip install face-recognition

比较两张图片

import cv2
import face_recognition

def find_face_encodings(image_path):
    # reading image
    image = cv2.imread(image_path)
    
    # get face encodings from the image
    face_enc = face_recognition.face_encodings(image)
    
    # return face encodings
    return face_enc[0]

# getting face encodings for first image
image_1 = find_face_encodings("image_1.png")

# getting face encodings for second image
image_2  = find_face_encodings("image_2.png")

# checking both images are same
is_same = face_recognition.compare_faces([image_1], image_2)[0]
print(f"Is Same: {is_same}")
if is_same:
    # finding the distance level between images
    distance = face_recognition.face_distance([image_1], image_2)
    distance = round(distance[0] * 100)
    
    # calcuating accuracy level between images
    accuracy = 100 - round(distance)
    
    print("The images are same")
    print(f"Accuracy Level: {accuracy}%")
else:
    print("The images are not same")

输出：

Is Same: True

The images are same

Accuracy Level: 70%