Face recognition in Python in AI

With the shocking performance of alphago last year, AI has once again become the darling of technology companies. AI involves many fields, and face recognition in image recognition is one of the interesting branches. Baidu's BFR, Face++'s open platform, Hanwang, iFlytek, etc. all provide face recognition APIs. For experienced programmers, they can write a small piece of code to see how many people are in a picture. There are no tall and tall people. on, just for fun, and only requires 7 lines of code. 　

　import cv2
　　face_patterns = cv2.CascadeClassifier('/usr/local/opt/opencv3/share/OpenCV/haarcascades/haarcascade_frontalface_default.xml')
　　sample_image = cv2.imread('/Users/abel/201612.jpg')
　　faces = face_patterns.detectMultiScale(sample_image,scaleFactor=1.1,minNeighbors=5,minSize=(100, 100))
　　for (x, y, w, h) in faces:
　　cv2.rectangle(sample_image, (x, y), (x+w, y+h), (0, 255, 0), 2)
　　cv2.imwrite('/Users/abel/201612_detected.png', sample_image);

Line 1 Introducing OpenCV

Open source is great, it allows us to broaden our horizons and does not have to reinvent the wheel. Instead of using PIL and combining it with a specific algorithm, OpenCV (http://opencv.org) is used directly. OpenCV is a cross-platform computer vision library released under the BSD license. It can run on Linux, Windows and Mac OS operating systems. It is lightweight and efficient. It is written in C/C++. It also provides Python, Ruby, MATLAB and other interfaces to implement Many general algorithms in image processing and computer vision.

Line 2 loads the classifier cv2.CascadeClassifier

CascadeClassifier is a cascade classifier used for face detection in Opencv. This class encapsulates the target detection mechanism, that is, the sliding window mechanism + Cascading classifiers. The data structure includes two main parts: Data and FeatureEvaluator. Data stores the classifier data loaded from the xml file obtained from training; while FeatureEvaluator is about the loading, storage and calculation of features. The training file used here is haarcascade frontalface default.xml provided by default in OpenCV. As for the specific principles of Haar and LBP, you can refer to the relevant documents of opencv. Simply, it can be understood as the feature data of the face.

Line 3 Load the target image imread

Face recognition systems are generally divided into: face image collection, face image preprocessing, face image feature extraction, and matching and recognition. For the sake of simplicity, read in the image,

Line 4 multi-scale detection detectMultiScale

Call the detectMultiScale function in CascadeClassifier for multi-scale detection, and the single-scale method will be called in multi-scale detection detectSingleScale. Parameter description:

scaleFactor is the scaling factor of the image

minNeighbors is the number of neighbors that should be retained for each cascade rectangle, which can be understood as how many faces are around a person

MinSize is the size of the detection window

These parameters can be adjusted for the picture, and the processing result returns a rectangular object list of faces.

Lines 5 and 6 draw a frame for each face

Loop through the list of rectangular objects of the face, obtain the coordinates, width and height of the face rectangle, and then draw it in the original picture To exit the rectangular frame, OpenCV's rectangle method is called, in which the color of the rectangular frame is adjustable.

Line 7 Save the test results

The mystery is not these 7 lines of code, but the related implementation in OpenCV. The Chinese website of OpenCV is also a good place to learn and experience.

Therefore, the 7 lines of code are just a gimmick, the real core is OpenCV. Then, there are some pitfalls when installing the OpenCV environment, so record them in particular.

Mac-based OpenCV environment

It is recommended to use Brew for installation. If brew is not installed, execute the following command first:

　$/usr/bin/ruby -e "$(curl -fsSL https://raw.githubusercontent.com/Homebrew/install/master/install)”

Then, specify the target warehouse $brew tap homebrew/science

Install OpenCV3 $brew install opencv3

The installation speed depends on the network. After installation, you need to bind the Python development environment. There are many ways:

　1) Add environment variables and add the site-packages of opencv to the PYTHONPATH

　2) Use ln soft connection to link cv2.so to the site-packages of the python environment

　3 )Directly cp cv2.so to the site-packages directory of the python environment

　More simply, execute the following command:

　echo /usr/local/opt/opencv3/lib/python2.7/site-packages >> /usr/local/lib/python2.7/site-packages/opencv3.pth
　　mkdir -p /Users/hecom/.local/lib/python2.7/site-packages
　　echo 'import site; site.addsitedir("/usr/local/lib/python2.7/site-packages")' >> /Users/hecom/.local/lib/python2.7/site-packages/homebrew.pth

The above is the detailed content of Face recognition in Python in AI. For more information, please follow other related articles on the PHP Chinese website!