song-gq / intelli-construct-flask

核酸截图OCR

Prerequisite: Conda,

If wants GPU enabled: Cuda, CuDNN Pytorch(with cuda)

Front-end: Vue2 https://github.com/Song-Gq/shanghai-nucleic-acid-ocr-vue

Back-end: Flask https://github.com/Song-Gq/shanghai-nucleic-acid-ocr

OCR-model: https://github.com/JaidedAI/EasyOCR

TODOLIST

• enable deployment mode
• client token
• disable excess file size and types
• export to excel interface
• delete expired excel files on server
• web table for needing attention samples
• recognition result roboustness problem
• running efficiency on cpu
• list.index() to fuzz_index()
• choose folder to upload
• batch delete
• enable HTTPS

依赖

$ conda create -n ocr-sample-flask python=3.6
$ conda activate ocr-sample-flask
$ pip install easyocr
$ pip install flask flask_cors xlwt

识别模型

将.EasyOCR解压缩并放至~/.EasyOCR

• For Windows

  C:\Users\${用户名}\.EasyOCR

• For Linux

  /home/${用户名}/.EasyOCR

共包含两个模型文件

• ~/.EasyOCR/model/craft_mlt_25k.pth

  https://github.com/JaidedAI/EasyOCR/releases/download/pre-v1.1.6/craft_mlt_25k.zip

• ~/.EasyOCR/model/zh_sim_g2.pth

  https://github.com/JaidedAI/EasyOCR/releases/download/v1.3/zh_sim_g2.zip

运行

$ cd ocr-sample-flask
$ python main.py

通过浏览器访问 http://127.0.0.1:5000

• 初次识别可能会自动下载模型，下载完成后可能会需要一些时间进行处理
• 输出excel文件在代码根目录下

部署

• 请自行配置Nginx，/etc/nginx/conf.d/

  server {
      listen 8888; #外部HTTP访问端口
      server_name localhost;
      location / {
          include uwsgi_params;
          uwsgi_pass 127.0.0.1:5000; #uwsgi端口
       }
  }

  HTTPS配置

  server {
      listen 443 ssl; #外部HTTPS访问端口
      server_name ocr.gqsong.xyz;
      ssl_certificate ocr.gqsong.xyz_bundle.crt;
      ssl_certificate_key ocr.gqsong.xyz.key;
      ssl_session_timeout 5m;
      ssl_protocols TLSv1.2 TLSv1.3;
      ssl_ciphers ECDHE-RSA-AES128-GCM-SHA256:HIGH:!aNULL:!MD5:!RC4:!DHE;
      ssl_prefer_server_ciphers on;
      location / {
          include uwsgi_params;
          uwsgi_pass 127.0.0.1:5000; #uwsgi端口
          client_max_body_size 20m;
       }
  }

• uwsgi配置文件已包含在代码根目录，端口5000

  [uwsgi]
  # 使用nginx时为socket而非http
  socket = 127.0.0.1:5000 
  # conda environment
  home=/home/veocw/anaconda3/envs/ocr-sample-flask
  # Flask script
  wsgi-file=/mnt/data/sgq/ocr/ocr-sample-flask/main.py
  callable=app
  # 根据GPU Memory调整Process数
  processes=2
  threads=16
  buffer-size=32768
  master=true
  stats=/mnt/data/sgq/ocr/ocr-sample-flask/uwsgi.status
  pidfile=/mnt/data/sgq/ocr/ocr-sample-flask/uwsgi.pid
  lazy=true

• 然后运行，通过浏览器访问http://127.0.0.1:8888/

$ sudo nginx
$ uwsgi config.ini
$ sudo nginx -s reload

(Optional) 生成可执行文件

$ pip install pyinstaller
$ pyinstaller -D main.py

uwsgi安装问题

使用如下命令修改源 sudo gedit /etc/apt/sources.list 在打开的文件中的最后两行加上如下代码，退出并保存： deb http://dk.archive.ubuntu.com/ubuntu/ xenial main deb http://dk.archive.ubuntu.com/ubuntu/ xenial universe

安装gcc和g+4.8

sudo update-alternatives --install /usr/bin/gcc gcc /usr/bin/gcc-4.8 100