Python has been hidden for ten years, and once image recognition is heard all over the world

Preface

Face processing is a hot topic in artificial intelligence , Face processing can automatically extract a large amount of information from the face using computer vision algorithms , For example, identity 、 Intention and emotion . Face plays an important role in visual communication , This is because the face contains a lot of nonverbal information , Therefore, face processing has always been a very interesting topic for computer vision learners , Because it involves different professional fields , For example, object detection 、 Feature point detection and object tracking . In this paper , We will first introduce the common face processing libraries , Then it explains how to use these databases for face detection , Lay a foundation for further related processing .\

Introduction to face processing

In order to focus on face processing related topics , We will use OpenCV library , as well as dlib、face_recognition and cvlib etc. Python package . meanwhile , This paper will use different methods in face processing , To introduce different methods to solve specific face processing tasks , The introduction of different methods will help you choose different methods according to different application needs . The following figure shows the related topics of face processing , And the available Python package ：\

As shown in the figure above , Face processing mainly includes the following topics ：

Face detection ： A special case of object detection , Its task is to find the position and size of all faces in the image . Facial feature point detection ： A special case of feature point detection , Its task is to locate the main feature points in the face . Face tracking ： A special case of object tracking , Its task is to use the additional information extracted in the continuous frames of the video , Find the location and size of all moving faces in the video . Face recognition ： A special case of target recognition , Its task is to recognize or verify a person from an image or video using the information extracted from the face ： Face recognition (1:N)： Find the closest match to the detected face in the known face set, face verification (1:1)： Check whether the detected face is the person it claims, as described above , Face processing mainly includes ： Face detection 、 Facial feature point detection 、 Face tracking and Face recognition , And in the daily recognition process , Use the most OpenCV、dlib、face_recognition and cvlib Database for face detection , And today we'll talk about opencv Application , This is also the simplest way , Because many algorithm modules have been encapsulated

Other face processing topics will be introduced in a subsequent series .

Okay , Don't talk much , On the subject

setup script

First step ： install opencv modular

use pip When the management tool installs library files , Default to use foreign source files , Therefore, the download speed in China will be relatively slow , Maybe it's just 50KB/s. fortunately , Some top scientific research institutions in China have prepared various images for us , Download up to 2MB/s

Can be used in pip When , Add parameters -i Image and address ( Such as

pypi.tuna.tsinghua.edu.cn/simple),

for example ：

pip install -i pypi.tuna.tsinghua.edu.cn/simple opencv-contrib-python

After installation , The following content will appear in your installation path \

These are the opencv Its own algorithm model , Can be called directly

The following is the code implementation of the call

Start with the simple , First recognize the face in the picture \

# opencvimport cv2
face_detector = cv2.CascadeClassifier('./haarcascade_frontalface_alt.xml')
img = cv2.imread('./image.jpeg')
face_zones = face_detector.detectMultiScale(img)
for x,y,w,h in face_zones:    
    cv2.circle(img,center=(x+w//2,y+h//2),radius=w//2,color=[0,0,255],thickness=2)
cv2.imshow('face',img)
cv2.waitKey(0)cv2.destroyAllWindows()
 Copy code 

Execution results , Just change the address of the picture

When the picture is linked , What is formed is the recognition effect of video \

Code up \

''' Turn on the camera '''

import numpy as np
import cv2

face_detector = cv2.CascadeClassifier('./haarcascade_frontalface_alt.xml')
video = cv2.VideoCapture(0)# Turn on the camera 
while True:
    flag,frame = video.read()
    if flag == False:
        break
    gray = cv2.cvtColor(frame,code = cv2.COLOR_BGR2GRAY)
    face_zones = face_detector.detectMultiScale(gray)
    for x,y,w,h in face_zones:
        cv2.circle(frame,center = (x+w//2,y+h//2),radius = w//2,color = [0,0,255],thickness = 2)
    cv2.imshow('ttnk',frame)
    key = cv2.waitKey(41)
    if key == ord('q'):# Exit conditions 
        break
cv2.destroyAllWindows()
video.release()
 Copy code 

Much of this is a dead cycle , Continuously update the obtained picture information in a circular way , Finally, integrate , Is the content of a video

And the acquisition of test images , You can see the crawler tutorial I wrote to you before , Code directly , Can pay attention to ：Java After the alliance of architects , The background to reply 【 Reptiles 】 Get oh