Recycling Classification Server

A Flask-based server that uses a TensorFlow Lite model to classify images of recyclable items into different categories. The server provides recycling instructions and examples for each category.

Features

• Image classification for 7 recycling categories:

  • Cardboard
  • Food Waste
  • Glass
  • Metal
  • Paper
  • Plastic
  • Other

• Real-time classification with confidence scores

• Detailed recycling instructions for each category

• Automatic uncertainty handling for low-confidence predictions

• CORS support for cross-origin requests

• Built with TensorFlow Lite for efficient inference

Requirements

• Python 3.8+
• TensorFlow Lite model file (recycling_model.tflite)

Installation

1. Clone the repository:

git clone <repository-url>
cd recycling-server

2. Create and activate a virtual environment (recommended):

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

3. Install dependencies:

pip install -r requirements.txt

4. Ensure the TensorFlow Lite model file (recycling_model.tflite) is in the root directory.

Project Structure

recycling-server/
├── app.py                 # Main server application
├── recycling_model.tflite # TFLite model file
├── requirements.txt       # Python dependencies
└── README.md             # This file

API Endpoints

GET /

Health check endpoint that returns server status and available categories.

Response:

{
    "status": "Recycling Classification Server is Running!",
    "model_loaded": true,
    "categories": ["Cardboard", "Food_Waste", "Glass", "Metal", "Paper", "Plastic", "Other"]
}

POST /predict

Endpoint for image classification.

Request:

• Method: POST
• Content-Type: multipart/form-data
• Body: image file (field name: "image")

Response:

{
    "success": true,
    "category": "Plastic",
    "confidence": 0.95,
    "instructions": [
        "Rinse clean",
        "Check recycling number",
        "Remove caps and labels",
        "Crush to save space"
    ],
    "examples": "Bottles, containers, packaging"
}

Configuration

The server includes several configurable parameters:

• CONFIDENCE_THRESHOLD: Minimum confidence level for predictions (default: 0.70)
• CATEGORIES: List of classification categories
• INSTRUCTIONS: Dictionary of recycling instructions and examples for each category

Running the Server

Development

python app.py

The server will start on http://localhost:5000

Production

gunicorn app:app

Error Handling

The server includes comprehensive error handling:

• Image processing errors
• Model loading failures
• Invalid requests
• Low confidence predictions

For uncertain classifications (confidence < 70% or close predictions), the server returns detailed feedback and improvement suggestions.

Model Information

The server uses an Xception-based TensorFlow Lite model optimized for recycling classification. The model:

• Accepts 299x299 RGB images
• Provides confidence scores for 7 categories
• Is optimized for mobile and edge deployment
• Accuracy: 97%

Security Notes

• The server includes CORS support for cross-origin requests
• Input validation is implemented for file uploads
• Error messages are sanitized for production use