README.md

metadata

title: OCR IITRoorkie
emoji: 📚
colorFrom: indigo
colorTo: green
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference

OCR and Keyword Search Web Application

This web application performs Optical Character Recognition (OCR) on uploaded images containing text in both Hindi and English, and provides a keyword search functionality.

Setup

1. Install the required dependencies: pip install -r requirements.txt This contains crucial libraries like transformers, gradio, pillow, tesseract, pytesseract

2. Install Tesseract OCR: For Windows, Download and install from https://github.com/UB-Mannheim/tesseract/wiki

3. Update the tesseract path in script (this was not needed while deploying to Hugging Face Space but had to use it while running it locally on my machine)

Running Locally

To run the application locally: python app.py

Deployment

To deploy on Hugging Face Spaces:

1. Created a new Space on Hugging Face.
2. While creating space, I set the Space SDK to Gradio
3. Upload the app.py file and created requirements.txt and packages.txt for libraries and packages respectively

Usage

1. Upload an image containing Hindi and English texts.
2. Enter a keyword to search within the extracted text.
3. The application will display the extracted text and search results.

Note: The OCR accuracy may vary depending on the image quality. Might get incorrect readings if the image has hazy words.