yunjuanwang/ocr-from-scratch

Implementation of OCR from scratch using Python and TensorFlow

OCR-from-scratch

Implementation of OCR (Optical Character Recognition) from scratch using Python and TensorFlow

Authors

Usama Muneeb (GitHub); Yunjuan Wang (GitHub)

University of Illinois at Chicago (UIC)

1. Running the Server

1.1 Setting up the environment

1.1.1 Install Node.js.

The application is developed on version 10 LTS. Other versions may continue to support but are untested.

Install the required Node packages:

Clone the Git repository, enter the directory and install packages:

git clone https://github.com/usamamuneeb/OCR-from-scratch.git
cd OCR-from-scratch
npm install formidable socketio gm pdfkit

Install the GraphicsMagick DLL for the gm Node package.

Windows Download

1.1.2 Install Python

Python 3.6.X is required. As of yet, TensorFlow works with Python 3.6.X.

Python Downloads

Install respective Python packages

python -m pip install tensorflow opencv image numpy

1.2 Running the Server

Make sure you are in the project directory and run:

cd OCR-from-scratch
node server.js

2. Accessing the frontend

The frontend can be accessed at port 8000 of the machine running the server application.

If using on local system, open a browser and point to:

http://localhost:8000