Abstract

This project introduces a robust and scalable API for intelligent text extraction from images, leveraging Azure Computer Vision OCR and FastAPI. Designed to streamline document digitization and improve accessibility, the system supports various image formats, automatically optimizes images for processing, and delivers precise text recognition. By integrating advanced cloud-based OCR capabilities with an intuitive API framework, this project simplifies text extraction workflows for developers and businesses. Key features include automatic resizing, format conversion, and seamless deployment via the Render platform, ensuring high performance and usability. This innovative solution demonstrates the potential of combining Azure’s cognitive services with FastAPI to create efficient and user-friendly tools for text extraction and analysis.

Methodology

This project integrates Azure Computer Vision OCR with FastAPI to deliver an efficient and user-friendly API for image-to-text extraction. The methodology encompasses the following key components:

1. Image Preprocessing

Format Support: The API supports JPEG and PNG formats. Unsupported image formats are automatically converted to a compatible format using the Pillow library.
Automatic Resizing: Very large images are resized to optimize processing time and accuracy without compromising the quality of the extracted text.

2. Azure OCR Integration

The core of the system is powered by Azure Computer Vision OCR, a cloud-based solution offering fast and accurate text recognition.
The API communicates with Azure’s endpoint to process images and retrieve the extracted text.

3. API Design and Development

FastAPI Framework: The API is developed using FastAPI for high performance and ease of use.
Endpoints:
Root Endpoint: Provides a status check to ensure the API is running correctly.
Text Extraction Endpoint: Allows users to upload images and receive the extracted text in JSON format.

4. Error Handling

Invalid Input Formats: Returns clear error messages if the uploaded file is not a supported image format.
Azure Service Issues: Provides descriptive error messages when there is an issue with Azure’s OCR service.
Internal Server Errors: Ensures general server-side issues are logged and reported effectively.

5. Deployment

The API is hosted on the Render platform, ensuring reliability and scalability for production use.
Seamless deployment allows users to access the service without needing extensive technical setup.

Installation and Usage

Clone the Repository

To start using the Azure OCR API, clone the repository from GitHub using the following command:

Install Dependencies

Install the required dependencies using the requirements.txt file:

Set Up Configuration

Create a config.py file and add your Azure OCR credentials:

To use the Azure OCR API, you need to obtain an Azure Computer Vision subscription key and endpoint. Follow these steps:

Create an Azure Account

If you don’t already have an Azure account, create one at Azure Portal.

Create a Computer Vision Resource

• Go to the Azure Portal and navigate to the Create a resource section.

• Search for Computer Vision and select Create.

• Fill in the required details (e.g., subscription, resource group, region) and create the resource.
  Obtain API Key and Endpoint

• Once the resource is created, go to the Keys and Endpoint section under the resource overview.

• Copy your Endpoint URL and one of the API Keys.

Set Up Your Configuration File

• Create a config.py file in the root directory of your project.

• Add the below lines to store your credentials:

Azure OCR Documentation

To understand the capabilities and limitations of the Azure OCR API, visit the official Azure Computer Vision Documentation.

Run the API

Start the FastAPI server locally with the following command:

• Access the API documentation at http://127.0.0.1:8000/docs.

Test the API

Open the Swagger UI by navigating to the local server URL:

http://127.0.0.1:8000/docs

• Locate the /extract-text/ endpoint and click Try it out.

• Upload an image under the file parameter and click Execute.

• View the JSON response in the output section.

Deployed Version

You can access the deployed version of this API here:

Render Deploy Link

API Overview: GET and POST Methods

• This image shows the GET and POST methods available in the Azure OCR API. The GET method is used to check if the API is running, while the POST method is the main function for extracting text from images. Open the POST section to test the API with an image.

Using Try It Out for Image Testing

• Click on the Try it out button under the POST method. Here, you can upload an image from your local computer. Please ensure the image is in JPEG or PNG format and is smaller than 5 MB. To upload an image, click on the Choose File button.

Executing the Process: Execute Button

• After uploading your image, click the Execute button to start the process. This triggers the API to analyze the image and extract the text. Wait for the response, which will appear in the Response Body section below.

Reviewing the Output: Response Body and JSON Download

• In this image, you can see the extracted text from an uploaded image, displayed in JSON format in the Response Body section. The example shows the table of contents of a machine learning course. You can download the extracted text as a JSON file by clicking the Download button.

Process Status

• A response code of 200 indicates the process was completed successfully. Other response codes are as follows:

• 400 Bad Request: An invalid request was made (e.g., an unsupported image format was uploaded).

• 401 Unauthorized: The Azure API key is invalid or missing.

• 500 Internal Server Error: An error occurred on the server.

For more details and to access the source code of this project, visit my GitHub Repository.