NodeJS OCR(Optical character recognition) app

Mulkiya-OCR is an experiment app, I’ve used Tesseract Open Source OCR engine to read text from images(for example mulkiya image) so that data input process can be automated and accurate as much as possible. The code is open source at github.

For those who are not familiar with word “Mulkiya”. It’s an arabic word which means vehicle registration card or vehicle ownership certificate.

Getting Started

• Install tesseract-ocr open source OCR engine
• git clone [email protected]:minhajkk/mulkiya-ocr.git clone repo from github
• npm install Install the dependencies in the local node_modules folder.
• bower instalk Install client side bower dependencies.
• npm start to run the application.

Demo

Structure

├── public (angularjs client side app)
│   ├── css 
│   ├── images
│   └── scripts (angularjs services/controllers sciprts)
├── server (nodejs server side app)
│   └── app.js (Entry point to run the application)
|   └── routes.js (routes file)
├── .buildpacks (heroku - installing nodejs/apt/tesseract multiple buildpacks)
├── .gitignore (git ignore)
├── Aptfile (heroku - for installing tesseract-ocr using apt-get)
├── bower.json (client side dependencies)
├── package.json (server side dependencies)
└── Procfile (heroku - process file)

Stack

The application is built upon nodejs and angularjs frameworks, find bellow more details about stack.

Server Side Dependencies (NPM)

• multer Multer is a node.js middleware for handling multipart/form-data.
• expressjs Web application framework.
• node-tesseract A simple wrapper for the Tesseract OCR package for node.js

Client Side Dependencies (Bower)

• angularjs Superheroic JavaScript MVW Framework.
• ui-routr For robust routing provider.
• angular-material For amazing material design support.
• ng-file-upload Amazing angularjs file uploading directive additionally allows mobile devices to capture using camera.

Deployment

Mulkiya-ocr is deployed on heroku @ https://mulkiya-ocr.herokuapp.com/. There’s a great step by step tutorial provided by heroku for deploying a Node.js app in minutes have a look in case you’re interested in using heroku.

Deployment was the main challenge for this application since I had to install tesseract-ocr on server before deployment but thanks for heroku’s multiple buildpack support It was great experiment using heroku multiple buildpacks on this app.

Some useful links

• Using Multiple Buildpacks for an App
• heroku-buildpack-tesseract
• Implementing OCR for Heroku Rails app
• Use multiple buildpacks on your app

What’s next?

1. Manipulate image before processing for example increase/reduce brightness/contrast so text can be extracted easily.
2. Develop algorithm’s to extract right information from mulkiya such as model, car, year etc etc