MV Bank Receipt OCR

An OCR server for reading bank maldivian bank receipts. A live production build is running on mvbankreceipt.xyz and a demo web app is running here.

Table of Contents

• Introduction
  • Endpoints
• Setting up for Development
  • Testing https Connection
• Deployment

Introduction

The project uses Tesseract OCR to convert bank receipt images into JSON. Receipts such as the one below

is expected to have an out put of

{
  "Amount": "2000",
  "Currency": "MVR",
  "Date": "20/01/20 00:00",
  "From": "Myself",
  "Message": "Still Waiting for Something to Happen",
  "Ref #": "1234356789ABCDEF",
  "Remarks": "This payment was never made",
  "Status": "Awaiting",
  "To": {
    "Account": "12345678910",
    "Name": "Someone you Know"
  }
}

Endpoints

The only endpoint is POST /ocr, and requires a file field, with an image file in .png, .jpg or .jpeg extension

Folder Structure

my-app/
  README.md
  requirements.txt
  vevn/
  src/
    ocr/
        preprocess/
        text_processing/
        ocr.py
    responses.py
    temp_clearer.py
    utils.py
  cert/

Setting up for Development

1. Clone the project
2. Create a virtual environment (option): python -m venv venv
3. Install dependencies: pip install -r requirements.txt
4. Run locally: python server.py

Testing https Connection

you can create self-signed SSL certificates to test the https connections using:

mkdir cert

openssl req -x509 -newkey rsa:4096 -keyout cert/key.pem -out cert/cert.pem -days 365

server.py will check for ssl certificates in the folder

Deployment

Currently, the script only runs in production mode if there is are SSL certificates available. This is reinforced to make sure sensitive information transmitted to the server is encrypted. It should also be noted that Flask is not the best package for creating a production-ready server.

To deploy, set FLASK_ENV environment variable to production:

export FLASK_ENV=production

then run python as a background process:

nohup python3 server.py > output.log &