使用阿里云人脸检测定位API

最新推荐文章于 2022-12-28 01:34:14 发布

weixin_34014555

最新推荐文章于 2022-12-28 01:34:14 发布

阅读量258

点赞数

文章标签： json python 人工智能

原文链接：https://my.oschina.net/wuliong/blog/1550613

版权

2019独角兽企业重金招聘Python工程师标准>>>

开通人脸识别服务，获取 AccessId 和 AccessSecret

在网上找一张照片，比如这张哈里王子出席晚宴的合影，记下图片url

新建文件夹 demo，进入文件夹，使用 wget 下载照片到文件夹内，并重命名为容易记忆的名字，比如 a.jpg，并从 github 下载 haarcascade_frontalface_default.xml

新建脚本 demo.py （这里使用的是 Python 3.6)，添加

import requests
import hmac
import hashlib
import datetime
import json 
import base64
from urllib.parse import urlparse
import cv2

然后是各种参数

image_url = "<Image URL>"
imagePath = "a.jpg"
url = "https://dtplus-cn-shanghai.data.aliyuncs.com/face/attribute"
ak_id = "<Your Access Id>"
ak_secret = "<Your Access Secret>"

定义几个函数

def to_md5_base64(str):
    hash = hashlib.md5()
    hash.update(str.encode("utf-8"))
    return base64.b64encode(hash.digest()).decode("utf-8").strip()

def to_sha1_base64(stringToSign, secret):
    hmacsha1 = hmac.new(secret.encode("utf-8"), stringToSign.encode("utf-8"), hashlib.sha1)
    return base64.b64encode(hmacsha1.digest()).decode("utf-8")

def get_current_date():
    date = datetime.datetime.strftime(datetime.datetime.utcnow(), "%a, %d %b %Y %H:%M:%S GMT")
    return date

发起请求所用的正文和headers

body = {"image_url":image_url,"type":0}

options = {
    "url": url,
    "method": "POST",
    "body": json.dumps(body, separators=(",",":")),
    "headers": {
        "Accept":"application/json",
        "Content-Type":"application/json",
        "Date": get_current_date(),
        "Authorization":""
    }
}

准备阿里云API校验所需的数字签名

urlPath = urlparse(options['url'])
if urlPath.query != '':
    urlPath = urlPath.path + "?" + urlPath.query
else:
    urlPath = urlPath.path

bodymd5 = to_md5_base64(options['body'])

stringToSign = options['method'] + '\n' \
             + options['headers']['Accept'] + '\n' \
             + bodymd5 + '\n' \
             + options['headers']['Content-Type'] + '\n' \
             + options['headers']['Date'] + '\n' \
             + urlPath

signature = to_sha1_base64(stringToSign, ak_secret)
authHeader = 'Dataplus '+ ak_id + ':' + signature 
options['headers']['authorization'] = authHeader

最后发起请求调用API

r = requests.post(url, data=options['body'], headers=options['headers'])
result = json.loads(r.text)

返回结果格式说明可以看这里

处理结果，将每张人脸的矩形框、特征点、姿态和瞳孔坐标整合成json格式数据。

faces = []
landmark_num = result['landmark_num']

for i in range(result['face_num']):
    face = {}
    face['rect'] = result['face_rect'][i*4:(i+1)*4]
    face['prob'] = result['face_prob'][i]
    face['pose'] = result['pose'][i*3:(i+1)*3]
    face['landmark'] = result['landmark'][i*landmark_num*2:(i+1)*landmark_num*2]
    face['iris'] = result['iris'][i*6:(i+1)*6]    
    faces.append(face)

最后用OpenCV展现照片，同时画出各种特征点以及瞳孔。

image = cv2.imread(imagePath)
for face in faces:
    [x, y, w, h] = face['rect']
    cv2.rectangle(image, (x,y), (x+w,y+h), (0,255,0), 2)
    lm = iter(face['landmark'])
    for x in lm:
        cv2.circle(image, (int(x),int(next(lm))), 2, (0,255,0), 2)
    iris = iter(face['iris'])
    for x in iris:
        cv2.circle(image, (int(x),int(next(iris))), int(next(iris)), (0,255,0), 1)

cv2.imshow("faces",image)
cv2.waitKey(1)

转载于:https://my.oschina.net/wuliong/blog/1550613