Home  >  Article  >  Backend Development  >  How to call Sogou AI API under python3

How to call Sogou AI API under python3

php中世界最好的语言
php中世界最好的语言Original
2018-04-12 09:58:542375browse

This time I will bring you the method of calling Sogou AI API under python3. What are the precautions for calling Sogou AI API under python3. The following is a practical case, let's take a look.

1. Background

a. Sogou has also released its own artificial intelligence API, including ID card OCR, business card OCR, text translation and other APIs. The initial test feels that the accuracy is so so.

b, based on python3.

c. It also has its own signature generation. With the foundation of Goose Factory, it is relatively simple to write.

d. However, Sougou is obviously not as good as Goose Factory in the standardization of interface. The main structure of different api response packets is actually inconsistent, so the implementation only has a simple structure...

2. Implementation code

Just put the code directly, it is also available on github: https://github.com/jdstkxx/PySougouAI

1, sogouai-example.py

# -*- coding: utf-8 -*-
'''
create by : joshua zou
create date : 2018.4.9
Purpose: check sougou ai api
'''
import glob,os
from SougouAPIMsg import *
#改成你自己搜狗AI的APPID、APPKEY、SecretKey
AppID = '0000'
ApiKey = '*********'
SecretKey= '0PLvS-AHShmq**************'
if name == "main":
  sg = SougouAPIMsg(AppID,ApiKey,SecretKey)
  for file in glob.glob('D:\python\*.jpg'):
    filename=os.path.split(file)[1].split('.')[0]
    #调用ocr识别
    apiname = 'ocr'
    rest =sg.apiSougouOcr(apiname,file)
    #调用身份证识别
    #rest =sg.apiSougouOcr('idcard',file)
    
    js= rest.json()
    retext =""
    if apiname=='ocr':
      #文字识别,rest应答包,字符串
      #成功 {"result":[{"content":"01245177\n","frame":["0,0","207,0","207,59","0,59"]}],"success":1}
      #失败 {"success":0}      
      if js['success']==1 :
        retext = js['result'][0]['content'].strip()         
    elif apiname == 'idcard':
      #身份证识别应答包,逼死强迫症啊,请求结构,应答结构都不一样
      '''
      {
      "result": {
      "住址": "xxxxxx",
      "公民身份号码": "11001xxx30",
      "出生": "19900101",
      "姓名": "xxXX",
      "性别": "X",
      "民族": "xxx"
      },
      "status": 0,
      "statusText": "Success"
      }
      '''
      if js['status']==0 :
        retext = js['result']['公民身份号码'].strip()      
    print(filename,retext)

2, SougouAPI.py

# -*- coding: utf-8 -*-
# 搜狗API字典
SougouAPI={
  #基本文本分析API
  "ocr":  {
    'APINAME':'图像识别', #API中文简称
    'APIDESC': '识别图像中的文字', #API描述
    'APIURL': 'http://api.ai.sogou.com/pub/ocr' #API请求URL
  },
  "idcard":{
    'APINAME':'身份证识别', #API中文简称
    'APIDESC': '身份证识别', #API描述
    'APIURL': 'http://api.ai.sogou.com/pub/ocr/idcard' #API请求URL
  },
  
}

3, SougouAPIMsg.py

# -*- coding: utf-8 -*-
'''
create by : joshua zou
create date : 2018.4.9
Purpose: check sougou ai api
'''
import requests
import base64
import hashlib
import hmac
import time
from urllib import parse
import json
from SougouAPI import *
class SougouAPIMsg(object):
  def init(self,AppID=None,ApiKey=None,SecretKey=None):
    if not AppID: AppID = '88888'
    if not ApiKey: ApiKey = '5ADwS88888888Dtr6QG2'
    if not SecretKey: SecretKey= '0PLvS-AH8888888889n6NF6fVVTt7m'
    self.app_id= AppID 
    self.app_key= ApiKey 
    self.app_secret= SecretKey 
  
  def get_time_stamp(self):
    return str(int(time.time()))
  
  '''
  1、应用相关前缀 {AuthPrefix}
  {AuthPrefix}=sac-auth-v1/{accessKey}/{secondsSinceEpoch}/{expirationPeriodInSeconds}
  2、请求相关数据 {Data}    
  {Data}={REQUEST_METHOD} + "\n" + {HOST} + "\n" + {URI} + "\n" + {SORTED_QUERY_STRING}
  其中,REQUEST_METHOD 为请求使用的 HTTP 方法, 如: GET|POST|PUT|DELETE
  HOST 为服务使用的域名, 如: api.ai.sogou.com
  URI 为请求的服务路径, 如: /speech/asr
  SORTED_QUERY_STRING 把 URL 中的 Query String(即 URL 中 “?” 后面的 “k1=v1&k2=v2” 字符串)进行编码后的结果。    
  编码方法为:
  将 Query String 根据 & 拆开成若干项,对每一项转换为 UriEncode(key) + "=" + UriEncode(value) 的形式, 其中 value 可以是空字符串
  将上面转换后的所有字符串按照字典顺序排序。
  将排序后的字符串按顺序用 & 符号链接起来。
  3、生成签名 {Signature}    
   {Signature}=HMAC-SHA256-BASE64({secretKey}, {AuthPrefix} + "\n" + {Data})
  4、生成认证信息, 通过 Authorization header 传递    
   Authorization: {AuthPrefix}/{Signature}
   Example:
   1\应用 accessKey/secretKey 分别为 bTkALtTB9x6GAxmFi9wetAGH / PMROwlieALT36qfdGClVz2iH4Sv8xZxe
    POST 方式访问 http://api.ai.sogou.com/speech/asr 接口
    GET 参数为 type=gbk&idx=1&starttime=1491810516
    当前系统时间为 1491810516
   2\计算过程     
    {AuthPrefix}="sac-auth-v1/bTkALtTB9x6GAxmFi9wetAGH/1491810516/3600"
    {Data}="POST\napi.ai.sogou.com\n/speech/asr\nidx=1&starttime=1491810516&type=gbk"
    {Signature}=HMAC-SHA256-BASE64("PMROwlieALT36qfdGClVz2iH4Sv8xZxe", {AuthPrefix} + "\n" + {Data})="vuVEkzcnUeFv8FxeWS50c7S0HaYH1QKgtIV5xrxDY/s="
   3\最终生成的 header 为
    Authorization: sac-auth-v1/bTkALtTB9x6GAxmFi9wetAGH/1491810516/3600/vuVEkzcnUeFv8FxeWS50c7S0HaYH1QKgtIV5xrxDY/s=
  '''
  def get_auth_sign_str(self,url,method):
    res= parse.urlparse(url)
    host= res.netloc
    uri = res.path
    query= res.query
        
    
    #1生成前置字符串
    authprefix= 'sac-auth-v1/%s/%s/%s' %(self.app_key,self.get_time_stamp(),3600)
    #2生成data
    query=dict( (k, v if len(v)>1 else v[0] )
            for k, v in parse.parse_qs(res.query).items() )     
    sort_dict= sorted(query.items(), key=lambda item:item[0], reverse = False)
    sortquerystr= parse.urlencode(sort_dict)
    data= '%s\n%s\n%s\n%s' %(method,host,uri,sortquerystr)
    #3生成signstr
    signstr ='%s\n%s' %(authprefix,data)
    #调用hamc.sha256
    shastr =hmac.new(self.app_secret.encode(), signstr.encode(), digestmod=hashlib.sha256).digest()
    #base64编码,还原成字符串
    signature = base64.b64encode(shastr).decode()
    
    #4组合成最终的授权码
    authstr= '%s/%s' %(authprefix,signature)
    return authstr
  '''
  $file = "OCR-test03.jpg";
  $url = "http://api.ai.sogou.com/pub/ocr";
  
  $hdr = array(
    "Content-Type: multipart/form-data",
    "Authorization: ".sign($ak, $sk, $url, "POST")
  ); // cURL headers for file uploading
  
  $postfields = array(
    "pic" => curl_file_create($file,'image/jpeg','a_b_c.jpg'),
  );
  
  $ch = curl_init();
  $options = array(
    CURLOPT_URL => $url,
    CURLOPT_HEADER => false,
    CURLOPT_POST => 1,
    CURLOPT_HTTPHEADER => $hdr,
    CURLOPT_POSTFIELDS => $postfields,
    CURLOPT_RETURNTRANSFER => true
  );
  '''
  def apiSougouOcr(self,apiname,picfilename):
    url = SougouAPI[apiname]['APIURL']
    name = SougouAPI[apiname]['APINAME']
    desc= SougouAPI[apiname]['APIDESC']
    
    authstr=self.get_auth_sign_str(url, method='POST')
    header={ "Authorization": authstr }
    
    picfile= {'pic':open(picfilename,'rb')}
    
    resp = requests.post(url,headers=header,files=picfile)      
    #print (resp.text)

I believe you have mastered the case after reading this article Method, for more exciting information, please pay attention to other related articles on the php Chinese website!

Recommended reading:

What are the methods for Dataframe query in pandas

selenium cookie skips the verification code login implementation step Detailed explanation

The above is the detailed content of How to call Sogou AI API under python3. For more information, please follow other related articles on the PHP Chinese website!

Statement:
The content of this article is voluntarily contributed by netizens, and the copyright belongs to the original author. This site does not assume corresponding legal responsibility. If you find any content suspected of plagiarism or infringement, please contact admin@php.cn