GreenSort - Intelligent Waste Classification System

GreenSort is a computer vision-based waste classification system designed to facilitate B2B recycling transactions. Powered by deep learning, GreenSort classifies waste materials from images, estimates their market value, and calculates shipping costs based on quantity and distance. The system is deployable as a desktop GUI or a web application using TensorFlow.js.

The model is trained using the RealWaste dataset and supports nine waste categories: Cardboard, Food Organics, Glass, Metal, Miscellaneous Trash, Paper, Plastic, Textile Trash, and Vegetation.

🚀 Key Features

Automated Waste Classification: Classifies waste types from images with >90% accuracy using MobileNetV2.
Recommendation System: Suggests waste handling actions based on type and weight using supervised learning.
Shipping Cost Calculation: Estimates shipping costs based on distance.
Interactive GUI: testing.py provides a user-friendly desktop interface for testing.
Web Deployment: TensorFlow.js models enable browser-based classification and pricing predictions.
Educational Add-On (Planned): Future integration of consultation and sustainability education features.

🗂️ Project Structure

GreenSort/
├── data/
│   └── realwaste-main/
│       ├── realwaste/                      # RealWaste dataset (download separately)
│       └── rekomendasi/                    # Recommendation dataset  
├── models/
│   ├── ComputerVision/
│   │   ├── greensort_model.h5              # Trained Keras classification model
│   │   └── tfjs_model/                     # TensorFlow.js models for browser deployment
│   ├── Sistemrekomendasi/
│       ├── models/                         # Trained Keras recommendation model
│       └── dataset.json                    # Dataset for recommendation
├── notebooks/
│   ├── CapstoneDBS_ComputerVision.ipynb    # Notebook for classification model training
│   └── CapstoneDBS_SistemRekomendasi.ipynb # Notebook for recommendation model training
├── src/
│   └── utils.py                            # Helper functions for preprocessing and evaluation
├── web/
│   ├── db/                                 # sql for database
│   ├── index.html                          # Web app entry point
│   ├── app.js                              # JavaScript for model inference
│   └── styles.css                          # Web app styling
├── testing.py                              # GUI tool for desktop testing
├── requirements.txt                        # Python dependencies
├── greensort_icon.ico                      # Optional app icon
├── LICENSE                                 # MIT License file
├── CHANGELOG.md                            # Version history and changes
└── README.md                               # Project documentation

✅ Prerequisites

Python 3.8 or higher (for training and GUI)
pip for Python dependency management
Node.js 16 or higher (for web deployment)
RealWaste Dataset from UCI Machine Learning Repository
GPU (Optional) for faster model training
Web Browser (e.g., Chrome, Firefox) for web deployment
Web Server (e.g., http-server or Live Server extension in VS Code) for local web testing

⚙️ Installation

For Desktop GUI

Clone the repository:

git clone https://github.com/yourusername/greensort.git
cd greensort

Create a virtual environment:

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

Install Python dependencies:

pip install -r requirements.txt

Example requirements.txt:

tensorflow>=2.12.0
tensorflowjs>=4.0.0
numpy>=1.24.0
matplotlib>=3.7.0
seaborn>=0.12.0
pillow>=9.5.0
scikit-learn>=1.2.0
pandas>=2.0.0
tk>=0.1.0           # For GUI (included with Python)

Download the RealWaste dataset and place it in data/realwaste/.
Prepare the trained models:
- Place greensort_model.h5 and recycling_recommendation_model.h5 in models/.
- Or train the models using notebooks/CapstoneDBS_ComputerVision.ipynb and notebooks/CapstoneDBS_SistemRekomendasi.ipynb.

For Web Deployment

Install Node.js (if not already installed):
- Download from nodejs.org or use a package manager:
```
sudo apt install nodejs npm  # On Ubuntu/Debian
```
Install TensorFlow.js:
```
npm install @tensorflow/tfjs
```

Convert Keras models to TensorFlow.js (if not already in models/tfjs_model/):

pip install tensorflowjs
tensorflowjs_converter --input_format keras models/greensort_model.h5 models/tfjs_model/classification
tensorflowjs_converter --input_format keras models/recycling_recommendation_model.h5 models/tfjs_model/recommendation

Set up a web server:
- Install a simple HTTP server:
```
npm install -g http-server
```
- Or use the Live Server extension in VS Code.

🔧 Usage

1. Training the Models

Open notebooks/CapstoneDBS_ComputerVision.ipynb for the classification model and notebooks/CapstoneDBS_SistemRekomendasi.ipynb for the recommendation model in Jupyter Notebook or JupyterLab.
Follow the steps to:
- Preprocess the dataset
- Train the models (MobileNetV2 for classification, neural network for pricing)
- Evaluate metrics (accuracy, MAE, R²)
Save the models to ../models/ComputerVision/greensort_model.h5 and ../models/Sistemrekomendasi/recycling_recommendation_model.h5.

Convert to TensorFlow.js:

import tensorflowjs as tfjs
tfjs.converters.save_keras_model(classification_model, 'models/tfjs_model/classification')
tfjs.converters.save_keras_model(recommendation_model, 'models/tfjs_model/recommendation')

2. Testing with the GUI

Run the desktop interface:
```
python testing.py
```
GUI features:
- Select Image: Upload a waste image.
- Set Quantity and Distance: Input weight (kg) and shipping distance (km).
- Classify Waste: View top prediction, confidence score, and estimated value.
- View Results: Displays classification and pricing details.
Example:
- Image: Plastic waste
- Weight: 2.5 kg
- Distance: 10 km
- Output: "Plastic" with 95% confidence → Estimated value: Rp 10,000 (based on Rp 4,000/kg)

3. Web Deployment

Create a web app in the web/ directory with the following files:
- index.html: Web interface
- app.js: JavaScript for model inference
- styles.css: Styling

Sample web/index.html:

<!DOCTYPE html>
<html lang="en">
<head>
    <meta charset="UTF-8">
    <meta name="viewport" content="width=device-width, initial-scale=1.0">
    <title>GreenSort - Waste Classification</title>
    <link rel="stylesheet" href="styles.css">
    <script src="https://cdn.jsdelivr.net/npm/@tensorflow/[email protected]/dist/tf.min.js"></script>
</head>
<body>
    <div class="container">
        <h1>GreenSort - Waste Classification</h1>
        <input type="file" id="imageInput" accept="image/*">
        <div>
            <label>Quantity (kg):</label>
            <input type="number" id="quantity" value="1.0" min="0.1" step="0.1">
        </div>
        <div>
            <label>Distance (km):</label>
            <input type="number" id="distance" value="10.0" min="0.1" step="0.1">
        </div>
        <button onclick="classifyImage()">Classify Waste</button>
        <div id="result"></div>
        <img id="preview" style="max-width: 400px; display: none;">
    </div>
    <script src="app.js"></script>
</body>
</html>

Sample web/app.js:

async function loadModels() {
    const classificationModel = await tf.loadLayersModel('/models/tfjs_model/classification/model.json');
    const pricingModel = await tf.loadLayersModel('/models/tfjs_model/pricing/model.json');
    return { classificationModel, pricingModel };
}

const CATEGORIES = ['Cardboard', 'Food_Organics', 'Glass', 'Metal', 'Miscellaneous_Trash', 'Paper', 'Plastic', 'Textile_Trash', 'Vegetation'];
const PRICE_PER_KG = {
    'Cardboard': 2000, 'Food_Organics': 500, 'Glass': 1500, 'Metal': 8000,
    'Miscellaneous_Trash': 300, 'Paper': 2500, 'Plastic': 4000, 'Textile_Trash': 1000, 'Vegetation': 600
};

async function classifyImage() {
    const imageInput = document.getElementById('imageInput');
    const quantity = parseFloat(document.getElementById('quantity').value);
    const distance = parseFloat(document.getElementById('distance').value);
    const resultDiv = document.getElementById('result');
    const preview = document.getElementById('preview');

    if (!imageInput.files[0]) {
        resultDiv.innerHTML = 'Please select an image.';
        return;
    }

    const { classificationModel, pricingModel } = await loadModels();

    // Load and preprocess image
    const img = new Image();
    img.src = URL.createObjectURL(imageInput.files[0]);
    img.onload = async () => {
        preview.src = img.src;
        preview.style.display = 'block';

        const tensor = tf.browser.fromPixels(img)
            .resizeNearestNeighbor([224, 224])
            .toFloat()
            .div(tf.scalar(255.0))
            .expandDims();

        // Classification
        const prediction = await classificationModel.predict(tensor).data();
        const predictedClassIndex = prediction.indexOf(Math.max(...prediction));
        const predictedClass = CATEGORIES[predictedClassIndex];
        const confidence = prediction[predictedClassIndex] * 100;

        // Pricing
        const pricePerUnit = PRICE_PER_KG[predictedClass];
        const pricingInput = tf.tensor2d([[...Array(9).fill(0).map((_, i) => i === predictedClassIndex ? 1 : 0), quantity, distance, pricePerUnit]]);
        const totalPrice = (await pricingModel.predict(pricingInput).data())[0];

        // Display results
        resultDiv.innerHTML = `
            <h3>Classification Result</h3>
            <p><strong>Type:</strong> ${predictedClass}</p>
            <p><strong>Confidence:</strong> ${confidence.toFixed(2)}%</p>
            <h3>Pricing Result</h3>
            <p><strong>Price per kg:</strong> Rp ${pricePerUnit.toLocaleString()}</p>
            <p><strong>Quantity:</strong> ${quantity.toFixed(2)} kg</p>
            <p><strong>Distance:</strong> ${distance.toFixed(2)} km</p>
            <p><strong>Total Value:</strong> Rp ${totalPrice.toLocaleString(undefined, {minimumFractionDigits: 0})}</p>
        `;
    };
}

Sample web/styles.css:

body {
    font-family: Arial, sans-serif;
    margin: 0;
    padding: 20px;
    background-color: #f0f0f0;
}
.container {
    max-width: 800px;
    margin: 0 auto;
    background: white;
    padding: 20px;
    border-radius: 8px;
    box-shadow: 0 0 10px rgba(0,0,0,0.1);
}
h1 {
    color: #27ae60;
}
input, button {
    margin: 10px 0;
    padding: 8px;
}
button {
    background-color: #27ae60;
    color: white;
    border: none;
    cursor: pointer;
}
button:hover {
    background-color: #219653;
}
#result {
    margin-top: 20px;
}

Run the web app:
```
http-server web/ -p 8080
```
- Open http://localhost:8080 in a browser.
- Upload an image, set quantity and distance, and click "Classify Waste" to see results.

4. Database

Create a database greensort.sql

Sample db/greensort.sql:

-- Table: waste_categories
CREATE TABLE waste_categories (
  category_id INT PRIMARY KEY AUTO_INCREMENT,
  category_name VARCHAR(50) NOT NULL UNIQUE,
  base_market_price DECIMAL(10,2) NOT NULL
);
   
-- Table: users (optional)
CREATE TABLE users (
  user_id INT PRIMARY KEY AUTO_INCREMENT,
  username VARCHAR(50) NOT NULL UNIQUE,
  email VARCHAR(100),
  created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP
);

-- Table: classifications
CREATE TABLE classifications (
  classification_id INT PRIMARY KEY AUTO_INCREMENT,
  image_path VARCHAR(255) NOT NULL,
  category_id INT NOT NULL,
  confidence DECIMAL(5,2) NOT NULL,
  classification_time TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
  user_id INT,
  FOREIGN KEY (category_id) REFERENCES waste_categories(category_id),
  FOREIGN KEY (user_id) REFERENCES users(user_id) ON DELETE SET NULL
);

-- Tabel referensi rekomendasi
CREATE TABLE recommendation_rules (
  rule_id INT PRIMARY KEY AUTO_INCREMENT,
  category_id INT NOT NULL,
  weight_min_kg DECIMAL(10,2) NOT NULL,
  weight_max_kg DECIMAL(10,2) NOT NULL,
  recomamendation TEXT NOT NULL,
  FOREIGN KEY (category_id) REFERENCES waste_categories(category_id)
);

-- Tabel log hasil rekomendasi
CREATE TABLE recommendations (
  recommendation_id INT PRIMARY KEY AUTO_INCREMENT,
  classification_id INT NOT NULL,
  estimated_weight DECIMAL(10,2) NOT NULL,
  generated_recommendation TEXT NOT NULL,
  created_at TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
  FOREIGN KEY (classification_id) REFERENCES classifications(classification_id)
);

♻️ Waste Categories & Pricing

Category	Price per kg (IDR)
Cardboard	2,000
Food Organics	500
Glass	1,500
Metal	8,000
Miscellaneous Trash	300
Paper	2,500
Plastic	4,000
Textile Trash	1,000
Vegetation	600

Note: Prices fluctuate based on market conditions. The pricing model accounts for a ±20% price variation.

📈 Model Performance

Image Classification Model (MobileNetV2)

Architecture: MobileNetV2 (fine-tuned)
Dataset: RealWaste (9 classes)
Metrics:
- Accuracy: 91.2%
- Precision (avg): 90.7%
- Recall (avg): 89.8%
- F1-Score (avg): 90.2%

Price Prediction Model

Architecture: Neural Network with Dropout
Dataset: Synthetic (1000 samples)
Metrics:
- MAE: ~Rp5,000
- R² Score: 0.95
- Relative Error: 2-3% of average price

Challenges:

Visual similarity between classes (e.g., Paper vs. Cardboard)
Market price fluctuations affecting pricing model accuracy

Suggestions:

Improve feature extraction
Use more advanced data augmentation
Incorporate real market data for price predictions

🔮 Future Improvements

Multi-label classification for multiple waste types in one image
Real-time shipping integration via APIs (e.g., Google Maps)
Chatbot assistant for education and user support
Model optimization for edge devices (quantization, pruning)
Dataset expansion to increase robustness
Real market data integration for more accurate price predictions
Enhanced web UI with real-time image previews and advanced visualizations

🤝 Contributing

We welcome contributions!

Fork the repository
Create a feature branch (git checkout -b feature/your-feature)
Commit changes (git commit -m "Add feature")
Push to GitHub (git push origin feature/your-feature)
Open a Pull Request

Please follow the Code of Conduct and PEP 8 style guidelines.

📦 CHANGELOG

All notable changes to this project are documented in the CHANGELOG.md file.

📄 License

This project is licensed under the MIT License

📬 Contact

Email: [email protected]
GitHub Issues: Open an issue

🙏 Acknowledgments

RealWaste Dataset — UCI Machine Learning Repository
TensorFlow
TensorFlow.js
Kementerian Lingkungan Hidup dan Kehutanan (KLHK) for waste management data

Name		Name	Last commit message	Last commit date
Latest commit History 37 Commits
${APPDATA}\npm		${APPDATA}\npm
.conda		.conda
data		data
dataTEST		dataTEST
mlruns		mlruns
models		models
node_modules		node_modules
notebooks		notebooks
src		src
web		web
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.MD		README.MD
cekversion.ipynb		cekversion.ipynb
cuda-keyring_1.1-1_all.deb		cuda-keyring_1.1-1_all.deb
cuda-keyring_1.1-1_all.deb.1		cuda-keyring_1.1-1_all.deb.1
package-lock.json		package-lock.json
package.json		package.json
requirements.txt		requirements.txt
tes.html		tes.html
testingAPI.py		testingAPI.py
testingLocal.py		testingLocal.py

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Repository files navigation

GreenSort - Intelligent Waste Classification System

🚀 Key Features

🗂️ Project Structure

✅ Prerequisites

⚙️ Installation

For Desktop GUI

For Web Deployment

🔧 Usage

1. Training the Models

2. Testing with the GUI

3. Web Deployment

4. Database

♻️ Waste Categories & Pricing

📈 Model Performance

Image Classification Model (MobileNetV2)

Price Prediction Model

🔮 Future Improvements

🤝 Contributing

📦 CHANGELOG

📄 License

📬 Contact

🙏 Acknowledgments

About

Uh oh!

Releases

Packages

Uh oh!

Contributors 2

Uh oh!

Languages

License

RazerArdi/GreenSort---Intelligent-Waste-Classification-System

Folders and files

Latest commit

History

Repository files navigation

GreenSort - Intelligent Waste Classification System

🚀 Key Features

🗂️ Project Structure

✅ Prerequisites

⚙️ Installation

For Desktop GUI

For Web Deployment

🔧 Usage

1. Training the Models

2. Testing with the GUI

3. Web Deployment

4. Database

♻️ Waste Categories & Pricing

📈 Model Performance

Image Classification Model (MobileNetV2)

Price Prediction Model

🔮 Future Improvements

🤝 Contributing

📦 CHANGELOG

📄 License

📬 Contact

🙏 Acknowledgments

About

Topics

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors 2

Uh oh!

Languages

Packages