人脸识别 face_recognition

Github开源人脸识别项目​​face_recognition​​，face_recognition是一个强大、简单、易上手的人脸识别开源项目

安装

pip install face_recognition

导入face_recognition模块

import  face_recognition

load_image_file 加载图像
调用face_recognition.load_image_file()读入图像。输出图像是rgb格式（opencv中是bgr格式）

import face_recognition
image = face_recognition.load_image_file("../datas/obama.jpg")

face_locations 人脸检测
能定位所有图像中识别出的人脸位置信息，返回值是列表形式，列表中每一行是一张人脸的位置信息，包括[top, right, bottom, left]

import face_recognition
import  cv2
#加载图像文件
image = face_recognition.load_image_file("../datas/hz.png")
image2 = cv2.cvtColor(image,cv2.COLOR_BGR2RGB)
cv2.imshow("src",image2)

face_locations = face_recognition.face_locations(image)
# 循环找到的所有人脸
for face_location in face_locations:
        # 打印每张脸的位置信息
        top, right, bottom, left = face_location
        print("A face is located at pixel location Top: {}, Left: {}, Bottom: {}, Right: {}".format(top, left, bottom, right))
        # 指定人脸的位置信息，然后显示人脸图片
        # face_image = image2[top:bottom, left:right]
        # cv2.imshow("src2", face_image)
        # 矩形框
        cv2.rectangle(image2, (left, top), (right, bottom), (0, 0, 255), 2)
cv2.imshow("dest",image2)
cv2.waitKey(0)

face_landmarks() 识别人脸关键点
调用face_recognition.face_landmarks(image)可识别出人脸关键点信息，包括眼睛、鼻子、嘴巴和下巴等
返回值是包含面部特征字典的列表，列表中每一项对应一张人脸，包括nose_bridge、right_eyebrow、right_eye、chine、left_eyebrow、bottom_lip、nose_tip、top_lip、left_eye几个部分，每个部分包含若干个特征点(x,y)，总共有68个特征点。

import cv2
import face_recognition
#加载图像文件
image = face_recognition.load_image_file("images/ldh2.jpg")
image2 = cv2.cvtColor(image,cv2.COLOR_BGR2RGB)

face_landmarks_list = face_recognition.face_landmarks(image)

facial_features = [
        'chin',
        'left_eyebrow',
        'right_eyebrow',
        'nose_bridge',
        'nose_tip',
        'left_eye',
        'right_eye',
        'top_lip',
        'bottom_lip'
    ]
thickness = 2
point_size = 1
point_color = (0, 255, 0)  # BGR

for face_landmarks in face_landmarks_list:

    for facial_feature in  face_landmarks:
        points_list = face_landmarks[facial_feature]
        print("The {} in this face has the following points: {}".format(facial_feature, points_list))
        # 在图像中画出每个人脸特征！
        for point in points_list:
            cv2.circle(image2, point, point_size, point_color, thickness)

cv2.imshow("dest",image2)
cv2.waitKey(0)

face_encodings() 面部编码
获取每个图像文件中每个面部的面部编码,每张人脸是一个128维的向量

import face_recognition
#加载图像文件
image = face_recognition.load_image_file("images/ldh2.jpg")

face_encodings = face_recognition.face_encodings(image)
for face_encoding in face_encodings:
    print("face_encoding len = {} \nencoding:{}\n\n".format(len(face_encoding),face_encoding))

face_encoding len = 128 encoding:[-0.12857245 0.2251953 -0.05680346
-0.009356 -0.07961649 -0.01976449
-0.03006363 -0.2188953 0.18227896 -0.06380306 0.23052536 -0.00517259
-0.24484953 -0.1024491 -0.07127093 0.10542907 -0.15712184 -0.24606238
-0.04181587 0.01180026 0.09390671 0.01014723 0.03316356 0.08216273
-0.08543567 -0.27527595 -0.10761252 -0.08561324 0.09562894 -0.06388975
-0.07860459 -0.01040951 -0.21186014 -0.10162984 0.09300554 0.08929634
-0.01133628 -0.02549033 0.25640246 -0.03459882 -0.22288607 0.06690408
0.05515105 0.31292963 0.14994277 0.06461242 0.00211182 -0.18974975
0.11956067 -0.07477805 0.09382927 0.21616243 0.18040094 0.02733615
0.03574533 -0.20213988 -0.00134893 0.08189995 -0.1417242 -0.00179647
0.10290838 -0.06461908 -0.01702856 -0.11400557 0.20824584 0.04568703
-0.07567319 -0.24240926 0.12750657 -0.16503944 -0.10546687 0.14743967
-0.13195604 -0.13922986 -0.26378733 -0.00479815 0.42761648 0.13261758
-0.17136547 0.09223576 -0.04313177 -0.03386433 0.07099032 0.1085121
-0.10494502 0.06504983 -0.15515642 -0.00395807 0.2313281 -0.01442198
-0.08458654 0.14688034 -0.00217457 0.07351732 0.06130501 0.01452726
-0.10001963 0.01601498 -0.12459313 -0.06574792 0.00460516 -0.01154638
-0.00899489 0.13434561 -0.14506203 0.06423949 -0.01264461 0.03912276
-0.06142928 0.01014899 -0.07765882 -0.04298633 0.10122736 -0.29780281
0.25160897 0.09939459 0.02078611 0.08443137 0.10742732 0.06015015
-0.00302691 -0.00466454 -0.20120506 -0.10478879 0.12094288 -0.0389271
0.19183818 -0.00275789]

compare_faces() 人脸匹配
compare_faces()方法可匹配两个面部特征编码，利用两个向量的内积来衡量相似度，根据阈值确认是否是同一人脸
第一个参数给出一个面部编码列表（很多张脸）
第二个参数给出单个面部编码（一张脸）
默认参数：tolerance=0.6,可根据自己需求更改，tolerance越小匹配越严格
匹配成功的脸返回True，匹配失败的返回False，顺序与第一个参数中脸的顺序一致

matches ==face_recognition.compare_faces(known_face_encodings,face_encoding,tolerance=0.39)

import face_recognition
# 加载已知面孔的图片
known_image = face_recognition.load_image_file("images/yz1.jpg")

# 加载未知面孔的图片
unknown_image = face_recognition.load_image_file("images/yz2.jpg")

# 加载2张已知面孔的图片
ldh_encoding = face_recognition.face_encodings(known_image)[0]
unknown_encoding = face_recognition.face_encodings(unknown_image)[0]

# 计算未知图片与已知图片的距离
results = face_recognition.compare_faces([ldh_encoding], unknown_encoding,tolerance=0.39)[0]
print(results)