metadata

title: GOT OCR Web App
emoji: 📄
colorFrom: blue
colorTo: green
sdk: streamlit
sdk_version: 1.21.0
app_file: app.py
pinned: false

OCR Web Application

Project Overview

This is a web-based Optical Character Recognition (OCR) application built using Streamlit. The app supports both English and Hindi languages, allowing users to upload images and extract text using advanced OCR models.

Live Demo

You can access the live demo of the application at: https://huggingface.co/spaces/Trisandhya/GOT-OCR-WEB-APP

How the Application Works

Choose Language: Select either English or Hindi using the sidebar instructions.
Upload Image: Use the file uploader to input an image in JPG, PNG, or JPEG format.
Text Extraction: For English, the app uses the GOT OCR 2.0 model to extract text, while for Hindi, it leverages EasyOCR.
Keyword Search: After text extraction, you can search for specific keywords within the extracted text. Matching keywords will be highlighted, and any missing keywords will be displayed in a warning message.
Reset: If needed, reset the session and upload a new image to start over.

Installation and Setup

Prerequisites:

Python 3.8 or higher
Required libraries listed in requirements.txt

Installation Steps:

Clone the repository:

git clone https://github.com/Trisandhyadevi/OCR.git

Navigate to the project directory
```
 cd OCR
```
Install the required dependencies:
```
pip install -r requirements.txt
```
Run the application:
```
streamlit run app.py
```

Description

This web application facilitates the conversion of images to text using the GOT OCR 2.0 Model for English text extraction. While the GOT model excels in processing English content, fine-tuning it on a Hindi dataset is not feasible on a CPU. Therefore, for Hindi text extraction, we utilize EasyOCR, which provides effective performance for this language.

GOT OCR 2.0 Model Overview

The GOT OCR 2.0 Model is a state-of-the-art OCR system designed for accurate text extraction from images. Key features include:

Multi-task Learning: The model supports various tasks beyond OCR, including layout analysis and object detection, making it versatile for diverse text recognition needs.
End-to-End Pipeline: It efficiently processes entire images, identifying and extracting text without the need for additional preprocessing steps.

Note: Currently, the model does not support all languages. Fine-tuning is required for languages not included in the pre-trained model. For more information on fine-tuning, visit the GOT OCR 2.0 Fine-tuning Guide.

For more technical details about the model architecture and usage, visit the GOT OCR 2.0 Model Documentation.

Folder Structure

.
├── app.py                
├── requirements.txt       
└── README.md

Deployment

To deploy the application to a Hugging Face cloud platform

Use GitHub Actions: Set up GitHub Actions in your repository to automate the deployment process to Hugging Face Spaces.
Follow Documentation: For detailed instructions on setting up GitHub Actions for Hugging Face Spaces, refer to the Hugging Face Spaces GitHub Actions Documentation.

Contact

For any further queries or assistance, feel free to reach out via email at:[email protected]