FASTAPI Microservice to Extract Text from Images

Python & FastAPI Tutorial: Create an ai microservice to extract text from images

Objective

Learn how to deploy an AI microservice REST API endpoint using FastAPI, pytesseract, streamlit cloud Platform

Streamlit deployment

Checkout the application live here 🚀

Screenshot 📸

Project Implementation Outline

• Setup Requirements.txt
• Setup Environment
• Setup FastAPI App
• FastAPI & Jinja Templates
• FastAPI & PyTest
• FastAPI Git & pre-commit
• Deploy to DigitalOcean
• Deploy Docker App to DigitalOcean App Platform
• FastAPI Settings & Environment Variables & dotenv
• Handling File Uploads
• Automated Testing File Uploads
• Image Upload Validation & Tests
• Implementing Tesseract & pytesseract
• Authorization Headers
• Production Endpoint & Authorization Tests
• One-Click Deploy on DigitalOcean App Platform

Project Requirements

fastapi
gunicorn
uvicorn
jinja2
pytest
requests
pre-commit
python-dotenv
python-multipart
aiofiles
pillow

Working with Pre-Commit

1. First, pip install pre-commit
2. Make a .pre-commit-config.yaml file
3. In the Terminal use : pre-commit-install
4. Next run the following command : pre-commit run --all-files

Working with PyTest

1. Install pytest using pip
2. add pytest.ini file to exclude the directories pytest looks for testing code
3. Using pytest -s to stdout the test responses from endpoints as well.

Working with Tesseract OCR

Reference - Tesseract OCR GitHub

Check the PyPi package here

Use following command to install wrapper class for Google Tesseract OCR Engine:

pip install pytesseract

1. Install tesseract using windows installer available at: https://github.com/UB-Mannheim/tesseract/wiki

2. Note the tesseract path from the installation. Default installation path at the time of this edit was: C:\Users\USER\AppData\Local\Tesseract-OCR. It may change so please check the installation path.

3. pip install pytesseract

4. Set the tesseract path in the script before calling image_to_string:

pytesseract.pytesseract.tesseract_cmd = r'C:\Users\USER\AppData\Local\Tesseract-OCR\tesseract.exe'

Securing the Endpoint using Auth Tokens

1. Generate a random token using Python secrets library

import secrets

secrets.token_urlsafe(32)

2. Save the generated tokens in a .env file