使用python AI快速比对两张人脸图像及遇到的坑

时间:2023-03-20 Python 集中营人气:0

本篇文章的代码块的实现主要是为了能够快速的通过python第三方非标准库对比出两张人脸是否一样。

实现过程比较简单，但是第三方python依赖的安装过程较为曲折，下面是通过实践对比总结出来的能够支持的几个版本，避免大家踩坑。

python版本：3.6.8
dlib版本：19.7.0
face-recognition版本：0.1.10

开始之前，我们选择使用pip的方式对第三方的非标准库进行安装。

pip install cmake

pip install dlib==19.7.0

pip install face-recognition==0.1.10

pip install opencv-python

然后，将使用到的模块cv2/face-recognition两个模块导入到代码块中即可。

# OpenCV is a library of programming functions mainly aimed at real-time computer vision.
import cv2

# It's loading a pre-trained model that can detect faces in images.
import face_recognition

新建一个python函数get_face_encodings，用来获取人脸部分的编码，后面可以根据这个编码来进行人脸比对。

def get_face_encodings(image_path):
    """
    It takes an image path, loads the image, finds the faces in the image, and returns the 128-d face encodings for each
    face

    :param image_path: The path to the image to be processed
    """
    # It's loading a pre-trained model that can detect faces in images.
    image = cv2.imread(image_path)

    # It's converting the image from BGR to RGB.
    image_RGB = image[:, :, ::-1]

    image_face = face_recognition.face_locations(image_RGB)

    # It's taking the image and the face locations and returning the face encodings.
    face_env = face_recognition.face_encodings(image_RGB, image_face)

    # It's returning the first face encoding in the list.
    return face_env[0]

上述函数中注释都是通过Pycharm插件自动生成的，接下来我们直接调用get_face_encodings函数分别获取两个人脸的编码。

# It's taking the image and the face locations and returning the face encodings.
ima1 = get_face_encodings('03.jpg')

# It's taking the image and the face locations and returning the face encodings.
ima2 = get_face_encodings('05.jpg')

# It's taking the image and the face locations and returning the face encodings.
ima1 = get_face_encodings('03.jpg')

# It's taking the image and the face locations and returning the face encodings.
ima2 = get_face_encodings('05.jpg')

上面我们选择了两张附有人脸的图片，并且已经获取到了对应的人脸编码。接着使用compare_faces函数进行人脸比对。

# It's comparing the two face encodings and returning True if they match.
is_same = face_recognition.compare_faces([ima1], ima2, tolerance=0.3)[0]

print('人脸比对结果：{}'.format(is_same))

人脸比对结果：False

这个时候人脸比对结果已经出来了，False代表不一样。这里compare_faces有一个比较重要的参数就是tolerance=0.3，默认情况下是0.6。

tolerance参数的值越小的时候代表比对要求更加严格，因此这个参数的大小需要根据实际情况设置，它会直接影响整个比对过程的结果。

加载全部内容