在face_predict下新建face-compare.py，完成人脸对比。

人脸比较：将两张人脸图像进行对比

将两者之间的相似值进行打印

阈值为0.6，阈值越小，条件越苛刻

import cv2
import face_recognition

加载被比较的图像

source_image = face_recognition.load_image_file("Face_database/hyz/hyz.png")

加载测试图像

compare_image = face_recognition.load_image_file("Facedatabase/hyz/hyz
near.png")

获取人脸位置并做单人脸容错处理

source_locations = face_recognition.face_locations(source_image)
if len(source_locations) != 1:
print("注意：图像一只能有一张人脸哦！")
exit(0)

获取人脸位置并做单人脸容错处理

compare_locations = face_recognition.face_locations(compare_image)
if len(compare_locations) != 1:
print("注意：图像二只能有一张人脸哦！")
exit(0)

绘制图像一的人脸

for (top, right, bottom, left) in source_locations:
print(top, right, bottom, left)
cv2.rectangle(source_image, (left, top), (right, bottom), (0, 255, 0), 2)

绘制图像二的人脸

for (top, right, bottom, left) in compare_locations:
print(top, right, bottom, left)
cv2.rectangle(compare_image, (left, top), (right, bottom), (0, 255, 0), 2)

获取图像一的面部编码

source_face_encoding = face_recognition.face_encodings(source_image)[0]
source_encodings = [
source_face_encoding,
]

获取图像二的面部编码

compare_face_encoding = face_recognition.face_encodings(compare_image)[0]

显示两张得到人脸后的图像

source_image = source_image[:, :, ::-1]
cv2.imshow("image", source_image)
cv2.waitKey(0)
compare_image = compare_image[:, :, ::-1]
cv2.imshow("image", compare_image)
cv2.waitKey(0)

查看面部一与面部二的比较结果，阈值为0.6，阈值越小越苛刻

face_distances = face_recognition.compare_faces(source_encodings, compareface
encoding, 0.6)

输出结果

print("正常阈值为0.6时，测试图像是否与已知图像{}匹配!".format("是" if face_distances else "不是"))
将两张照片图像加载进来后，先判断照片中是否只有一张人脸。如果照片中多于一张人脸，提示并退出；如果两张照片中都只有一张人脸，那么就将人脸圈出来，圈出来的人脸进行编码后进行比对，阈值为0.6，小于该阈值，就认为两张图像属于一张人脸。