Building a Simple Celebrity Face Matching System Using Baidu AI API and MySQL in Python

The author was inspired by movie scenes that instantly identify a person from a photo and decided to build a small prototype that can match a given image against a database of celebrity faces.

The workflow consists of five main steps:

Collect celebrity data by crawling online sources for images and related information.

Normalize the data and store the image file names and metadata in a MySQL table.

Use Baidu's face‑recognition API to encode images in base64 and send them for similarity comparison.

Upload a target image, compare it with every entry in the database, and record the similarity scores.

Output the best match along with the corresponding celebrity information and similarity percentage.

The core Python script (shown below) implements these steps, handling database connections, web requests, image encoding, and result parsing.

<code># encoding:utf-8
import base64
import urllib
import urllib2
import simplejson as json
from  os import listdir
import MySQLdb
import sys
reload(sys)
sys.setdefaultencoding('utf-8')

def conmysql():
    conn= MySQLdb.connect(
        host='localhost',
        port =3306,
        user='root',
        passwd='123456',
        db ='xxnlove',
        charset='utf8'
        )
    return conn

'''
人脸比对接口
'''

def facecompar(image01,image02):
    matchUrl = "https://aip.baidubce.com/rest/2.0/face/v2/match"
    f = open(image01, 'rb')
    img1 = base64.b64encode(f.read())
    f = open(image02, 'rb')
    img2 = base64.b64encode(f.read())
    params = {"images": img1 + ',' + img2}
    params = urllib.urlencode(params)
    access_token = '24.1a060b87a0dfcab77317999d.25922220.1505832798.282335-10029360'
    matchUrl = matchUrl + "?access_token=" + access_token
    request = urllib2.Request(url=matchUrl, data=params)
    request.add_header('Content-Type', 'application/x-www-form-urlencoded')
    response = urllib2.urlopen(request)
    content = response.read()
    if content:
        content = json.loads(content)
        similar = content['result'][0]['score']
        return similar

def compare():
    similarlist=[]
    similardict={}
    for img in listdir('./star/'):
        similarvalue=facecompar('compar.jpg','./star/'+img)
        similarlist.append(similarvalue)
        similardict[similarvalue]=img
    return similarlist,similardict

if __name__=="__main__":
    similarlist,similardict=compare()
    similarkey=sorted(similarlist)[-1]
    starname=similardict[similarkey]
    conn = conmysql()
    cur = conn.cursor()
    sql="select * from face where iamge='%s'" % starname
    cur.execute(sql)
    results = cur.fetchall()
    print "一共对数据库进行比对了"+str(len(similarlist))+"条信息"
    for info in results:
        print "匹配到明星的信息："+ info[0],info[1],info[2],"相似度："+str(similarkey)
    conn.close()
</code>

Running the program on a test image matched the celebrity "曾轶可" with a similarity of 63.69%, and the script printed the total number of comparisons performed. The author notes that image size inconsistencies affect accuracy and that the core face‑matching logic relies on Baidu's API rather than a custom algorithm, prompting further study of data structures and algorithms.