Using Baidu Intelligent Cloud face detection interface to achieve photo quality detection

First create baidu account , Log in to Baidu AI Open platform , Enter the Developer Center , Access control , Select face recognition in products and services . Then create an application .

Created successfully API Key and Secret Key

Next, use the just obtained API Key and Secret Key obtain access_token.

To the authorized service address https://aip.baidubce.com/oauth/2.0/token Send a request （ Recommended POST）, And in URL With the following parameters ：

• grant_type： Necessary parameters , Fixed for client_credentials;
• client_id： Necessary parameters , Applied API Key;
• client_secret： Necessary parameters , Applied Secret Key;
  Use postman Simulation request received access_token, as follows ：
  

Test interface

Face detection interface

Ability introduction

• Face detection ： Detect the face in the picture and mark the location information .
• Face key points ： Show the key information of face , And 150 Key information .
• Face attribute value ： Display face attribute information , Such age 、 Gender, etc .
• Face quality information ： Return the occlusion of all parts of the face 、 light 、 Fuzzy 、 Completeness 、 Confidence, etc .

Call interface

stay postman Call interface in ,url Add parameters in access_token, The value is the one you just applied for token. Set up Header in Content-Type Parameter is application/json. And then in body Add parameters in image and image_type, Represents picture information and picture type respectively , As shown in the figure below ：

Here, due to not buying all the returned error messages .

python call

After receiving Baidu's free resources, try again .
Use python Call the face detection interface to detect the quality of face photos .

Use the first three indicators in the interface document to evaluate the photo quality , If all the above three conditions are met, the photo is considered qualified .
Next use python To achieve ：

# encoding:utf-8

import requests
import base64
import os
import json

'''  Face detection and attribute analysis  '''
#  obtain access_token
# client_id  Obtained for the official website AK, client_secret  Obtained for the official website SK
# host = 'https://aip.baidubce.com/oauth/2.0/token?grant_type=client_credentials&client_id=【】&client_secret=【】'
# response = requests.get(host)
# if response:
# print(response.json())

# api Address ：https://cloud.baidu.com/doc/FACE/s/yk37c1u4t?_=1628489078955

""" occlusion: left_eye : 0.6, # The threshold of occlusion of the left eye  right_eye : 0.6, # Right eye occlusion threshold  nose : 0.7, # The threshold at which the nose is obscured  mouth : 0.7, # The threshold of mouth occlusion  left_cheek : 0.8, # The threshold of occlusion of the left cheek  right_cheek : 0.8, # The threshold of occlusion of the right cheek  chin_contour : 0.6, # Jaw occlusion threshold  blur: Less than 0.7 illumination: Greater than 40 """

#  Check whether a picture is qualified 
def is_ok(img):
    request_url = "https://aip.baidubce.com/rest/2.0/face/v3/detect"
    
    params = {
    "image":img, "image_type":"BASE64", "face_field":"faceshape,facetype,age,quality,angle"}
    access_token = ' You got it yourself from the official website access_token'
    request_url = request_url + "?access_token=" + access_token
    headers = {
    'content-type': 'application/json'}
    response = requests.post(request_url, data=params, headers=headers)
    #  Because free resources have throughput limitations , Therefore, if the throughput is exceeded, an error will be reported , Output the error here . Of course, you can also sleep every time the program runs 1 second , So it won't exceed qps 了 .
    if response.json()['error_code'] != 0:
        print(response.json()['error_msg'])
        return False
    if response:
        res = response.json()['result']
    else:
        print(' No response ')

    res = res['face_list'][0]
    occlusion = res['quality']['occlusion']
    left_eye = occlusion['left_eye']
    right_eye = occlusion['right_eye']
    nose = occlusion['nose']
    mouth = occlusion['mouth']
    left_cheek = occlusion['left_cheek']
    right_cheek = occlusion['right_cheek']
    chin_contour = occlusion['chin_contour']
    blur = res['quality']['blur']
    illumination = res['quality']['illumination']


    if left_eye < 0.6 and right_eye < 0.6 and nose < 0.7 and mouth < 0.7 and left_cheek < 0.8 and right_cheek < 0.8 and chin_contour < 0.6 and blur < 0.7 and illumination > 40:
        return True
    else:
        return False 

if __name__=='__main__':

    path = ' The path of the folder where the photos are stored '
    files = os.listdir(path)
    for file in files:
        f = open(path + '/' + file, 'rb')
        img = base64.b64encode(f.read())
        if is_ok(img):
            print(' qualified ')
        else:
            print(' unqualified ')

Download some pictures from the Internet to test , As shown in the following nine pictures :

Two runs , Take a look at the results :