Optical Character Recognition (OCR) Service
This repository contains a Flask-based REST API for performing Optical Character Recognition (OCR) on uploaded images. It utilizes the Flask-RESTx library to define the API endpoints and manage request and response formats.
Getting Started
Prerequisites
- Python (>=3.6)
- Install required packages using
pip install -r requirements.txt
Running the API
- Clone this repository:
git clone https://github.com/idan1356/flask_OCR_Service.git - Navigate to the project directory:
cd ocr-service - Run the application:
python app.py
The API will be accessible at http://localhost:5000/api/decipher-text.
API Endpoints
The API documentation provides detailed information about the available endpoints, request and response formats, and example usage. You can access the documentation by visiting /api/docs in your browser.
Upload Image for OCR
Endpoint: POST /api/decipher-text
This endpoint allows you to upload an image for Optical Character Recognition.
Parameters:
user_image: The image file to be processed.langs: Comma-separated list of languages in the image (see language list in config file).get_processed_image: Whether the processed image (rectangles around sentences found by the model) should be returned.
Responses:
200 OK: OCR successful. Returns OCR results and optional processed image.400 Bad Request: OCR unsuccessful. Returns an error message.
Project Structure
The project is organized as follows:
app.py: The main Flask application file.api/: Contains the API definition and resources.api/components/: Contains parser and data model definitions.api/utils/: Contains utility functions for processing images and handling API logic.
OCR Model Caching
The OCR model used by the service is LRU (Least Recently Used) cached to improve performance. The caching is implemented using the functools.lru_cache decorator, to cache OCR models based on the languages used for processing.
Configuration
You can configure languages and other settings in the config file.
Contributing
Contributions are welcome! Please open issues or submit pull requests for bug fixes, features, or improvements.
License
This project is licensed under the MIT License.