Spaces:
Running
Running
| #!/usr/bin/python3.6 | |
| # -*- coding: utf-8 -*- | |
| # | |
| # Copyright (C) 2021 # | |
| # @Time : 2022/1/6 23:28 | |
| # @Author : sml2h3 | |
| # @Email : sml2h3@gmail.com | |
| # @File : test_api.py | |
| # @Software: PyCharm | |
| import base64 | |
| import json | |
| import requests | |
| print(' ') | |
| # ******************OCR识别部分开始****************** | |
| host = "http://127.0.0.1:9898" | |
| # 目标检测就把ocr改成det,其他相同 | |
| # 方式一 | |
| file = open(r'test.jpg', 'rb').read() | |
| # file = open(r'test_calc.png', 'rb').read() | |
| api_url = f"{host}/ocr/file" | |
| resp = requests.post(api_url, files={'image': file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/ocr/file/json" | |
| resp = requests.post(api_url, files={'image': file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/ocr/b64" | |
| resp = requests.post(api_url, data=base64.b64encode(file).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/ocr/b64/json" | |
| resp = requests.post(api_url, data=base64.b64encode(file).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/det/file" | |
| resp = requests.post(api_url, files={'image': file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/det/file/json" | |
| resp = requests.post(api_url, files={'image': file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| # 滑块识别 | |
| target_file = open(r'match_target.png', 'rb').read() | |
| bg_file = open(r'match_bg.png', 'rb').read() | |
| api_url = f"{host}/slide/match/file" | |
| resp = requests.post(api_url, files={'target_img': target_file, 'bg_img': bg_file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/match/file/json" | |
| resp = requests.post(api_url, files={'target_img': target_file, 'bg_img': bg_file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/match/b64" | |
| target_b64str = base64.b64encode(target_file).decode() | |
| bg_b64str = base64.b64encode(bg_file).decode() | |
| jsonstr = json.dumps({'target_img': target_b64str, 'bg_img': bg_b64str}) | |
| resp = requests.post(api_url, data=base64.b64encode(jsonstr.encode()).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/match/b64/json" | |
| resp = requests.post(api_url, data=base64.b64encode(jsonstr.encode()).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| target_file = open(r'compare_target.jpg', 'rb').read() | |
| bg_file = open(r'compare_bg.jpg', 'rb').read() | |
| api_url = f"{host}/slide/compare/file" | |
| resp = requests.post(api_url, files={'target_img': target_file, 'bg_img': bg_file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/compare/file/json" | |
| resp = requests.post(api_url, files={'target_img': target_file, 'bg_img': bg_file}) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/compare/b64" | |
| target_b64str = base64.b64encode(target_file).decode() | |
| bg_b64str = base64.b64encode(bg_file).decode() | |
| jsonstr = json.dumps({'target_img': target_b64str, 'bg_img': bg_b64str}) | |
| resp = requests.post(api_url, data=base64.b64encode(jsonstr.encode()).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| api_url = f"{host}/slide/compare/b64/json" | |
| resp = requests.post(api_url, data=base64.b64encode(jsonstr.encode()).decode()) | |
| print(f"{api_url=}, {resp.text=}") | |
| # 方式二 | |
| # 获取验证码图片 | |
| # headers = { | |
| # "Content-Type": "application/x-www-form-urlencoded; charset=UTF-8", | |
| # "User-Agent": "Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/86.0.4195.1 Safari/537.36" | |
| # } | |
| # resp = requests.get('https://data.gdcic.net/Dop/CheckCode.aspx?codemark=408.15173910730016', headers=headers, verify=False) | |
| # captcha_img = resp.content | |
| # | |
| # 识别 | |
| # resp = requests.post(api_url, files={'image': captcha_img}) | |
| # print('验证码结果', resp.text) | |
| # | |
| # # 保存验证码图片以供验证 | |
| # with open('captcha.jpg', 'wb') as f: | |
| # f.write(captcha_img) | |
| # ******************OCR识别部分开始****************** | |