🥗 Nutrition & Ingredients Parser

A lightweight FastAPI application that extracts ingredients, allergens, and nutrition information from product packaging images using EasyOCR if running locally. The application can also be installed as Lamdba on AWS. If deployed to AWS it will use AWS Rekognition for text extraction.

🚀 Features

• Ingredient Extraction: Automatically identifies and lists ingredients from a cropped image.
• Allergen Detection: Specifically highlights potential allergens (e.g., "Contains: Milk", "May contain: Nuts").
• Nutrition Parsing: Extracts nutrition facts for easy digitization.
• OCR Powered: Utilizes EasyOCR for robust, offline-capable text recognition.
• Web Interface: Simple, user-friendly Tailwind-CSS UI for uploading and viewing results.

🛠️ Setup

Prerequisites

• Python 3.9+
• [Optional] GPU for faster OCR processing (defaults to CPU)
• AWS CLI configured (if using Rekognition)

Local Development

1. Clone the repository:

   git clone <repository-url>
   cd ingredients_parser

2. Create and activate a virtual environment:

   python -m venv venv
   source venv/bin/activate  # On Windows use `venv\Scripts\activate`

3. Install dependencies:

   pip install -r backend/requirements.txt
   pip install pytest  # Required for running tests

4. Run the application locally:

   ./run_local.sh

   This will start the FastAPI backend on port 8000 and a simple web server for the frontend on port 3000.

5. Access the UI: Navigate to http://localhost:3000/ in your browser.

Switching OCR Provider

The backend supports both EasyOCR and AWS Rekognition. You can switch between them using the OCR_TYPE environment variable.

• EasyOCR (Default): export OCR_TYPE=easyocr
• AWS Rekognition: export OCR_TYPE=rekognition

Note: Using Rekognition requires boto3 to be configured with valid AWS credentials.

AWS Deployment (Terraform)

The infrastructure is managed using Terraform and includes an ECR repository, a Lambda function (with Function URL), an S3 bucket for the static frontend, and a CloudFront distribution with TLS (ACM) for secure, high-performance delivery. It uses a remote state stored in S3 for better collaboration and state persistence.

1. Prerequisites:

   • Docker installed and running.
   • AWS CLI configured with appropriate permissions.
   • You need to be logged in to your AWS Account and this has to be enabled in the session.
   • Terraform installed.
   • S3 Bucket for Remote State: An S3 bucket is required to store the Terraform state. By default, it expects de-grimmfrost-terraform-state.
   • Route 53 Hosted Zone: A hosted zone for your domain (e.g., grimmfrost.de) must already exist in your AWS account.

2. Initialize and Create ECR Repository:

   cd terraform
   terraform init
   terraform apply
   cd ..

   This will fail with the configuration of the Lambda. Run the deploy_lambda.sh once to have an initial image in the ECR Repository.

3. Build and Push the Lambda Image:

   ./deploy_lambda.sh

   Note: On the first run, the script will push the image to ECR but skip the Lambda code update since the function hasn't been created yet.

4. Deploy the Rest of the Infrastructure:

   cd terraform
   terraform apply
   cd ..

5. Configure and Deploy Frontend:

   • Get the Lambda URL from Terraform outputs: cd terraform && terraform output lambda_function_url
   • Update API_URL in frontend/index.html with this value.
   • The frontend is now accessible via the custom domain configured in Terraform (default: https://ingredients.grimmfrost.de).
   • Sync the frontend to S3:

   aws s3 sync frontend/ s3://$(cd terraform && terraform output -raw frontend_s3_bucket)/

   • GitHub Actions Note: After the first successful deployment, copy the cloudfront_distribution_id from Terraform outputs and update the CLOUDFRONT_DISTRIBUTION_ID in .github/workflows/ci.yml to enable automatic cache invalidation on future pushes.

6. Access the App: Use the frontend_s3_url (CloudFront URL with custom domain) from Terraform outputs to open the application in your browser.

Public Deployment on ingredients.grimmfrost.de

There is a Deployment available on https://ingredients.grimmfrost.de/. This is for Testing Purposes only and is protected by a user and password.

📁 Project Structure

• backend/: FastAPI application, OCR logic, and providers.
• frontend/: Static HTML/JS frontend, ready for S3 deployment.
• terraform/: Infrastructure as Code for usage on AWS.
• tests/: Test suites for OCR and parsing logic.

🧪 Testing

The project uses pytest for testing. To run the tests, use:

pytest tests/test_ingredients_vision.py

Note: If easyocr or torch are not installed, these tests will be skipped automatically.

Sample images for testing can be found in tests/static/cropped_ingredients.

📝 Usage Notes

• For best results, ensure the image is cropped to the relevant section of the product packaging (ingredients or nutrition table).
• The current version is optimized for English text.

---

Built with Python, Flask, and EasyOCR.