🗑️ Waste Classification Model - Training & Deployment

A comprehensive YOLOv8-based waste classification system that automatically identifies and categorizes waste materials into 5 main categories: Glass, Organic, Others, Packaged, and Plastic.

📋 Table of Contents

• Overview
• Model Categories
• Project Structure
• Quick Start
• Training
• Inference
• Web Application
• Category Management
• Troubleshooting
• Contributing

🎯 Overview

This project implements a state-of-the-art waste classification system using YOLOv8 (You Only Look Once version 8) for real-time object detection and classification. The system is designed to help improve waste management practices by automatically sorting waste materials.

Key Features

• ✅ Real-time Detection: Fast and accurate waste classification
• ✅ 5 Waste Categories: Glass, Organic, Others, Packaged, Plastic
• ✅ Web Interface: User-friendly Streamlit application
• ✅ Category Mapping: Intelligent mapping from model categories to desired categories
• ✅ Multiple Input Sources: Images, webcam, and video files
• ✅ Confidence Control: Adjustable detection sensitivity
• ✅ Detailed Results: Category counts and confidence scores

🏷️ Model Categories

The model classifies waste into the following categories:

Category	Description	Examples
Glass	Glass containers and bottles	Wine bottles, jars, glass containers
Organic	Biodegradable waste	Food scraps, leaves, organic matter
Others	Miscellaneous waste	Metal, ceramics, non-recyclable items
Packaged	Packaging materials	Cardboard, paper, packaging
Plastic	Plastic containers and items	Plastic bottles, bags, containers

Category Mapping

The model uses intelligent category mapping to convert from the original training categories to your desired categories:

BIODEGRADABLE → Organic
CARDBOARD     → Packaged
GLASS         → Glass
METAL         → Others
PAPER         → Packaged
PLASTIC       → Plastic

📁 Project Structure

Train/
├── Waste-Classification-using-YOLOv8/
│   ├── dataset/                          # Training and validation data
│   │   ├── train/images/                 # Training images
│   │   ├── train/labels/                 # Training labels
│   │   ├── val/images/                   # Validation images
│   │   └── val/labels/                   # Validation labels
│   ├── streamlit-detection-tracking - app/  # Web application
│   │   ├── app.py                        # Main Streamlit app
│   │   ├── helper.py                     # Helper functions
│   │   ├── settings.py                   # App configuration
│   │   ├── category_mapper.py            # Category mapping utility
│   │   └── weights/
│   │       └── best.pt                   # Trained model weights
│   ├── data.yaml                         # Dataset configuration
│   ├── category_mapper.py                # Category mapping utility
│   ├── verify_model_categories.py        # Model verification script
│   ├── test_inference.py                 # Inference testing script
│   ├── train_new_model.py                # Model training script
│   ├── quick_start.py                    # Quick start script
│   └── requirements.txt                  # Python dependencies
└── README.md                             # This file

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (recommended for training)
• 8GB+ RAM

Installation

1. Clone the repository (if not already done):

   git clone <repository-url>
   cd WasteNet/Train

2. Navigate to the project directory:

   cd Waste-Classification-using-YOLOv8

3. Install dependencies:

   pip install -r requirements.txt

4. Verify the model:

   python verify_model_categories.py

5. Launch the web application:

   python quick_start.py

Quick Test

Test the model on a sample image:

python test_inference.py --model "streamlit-detection-tracking - app/weights/best.pt" --image "dataset/train/images/0.jpg"

🏋️ Training

Dataset Preparation

The dataset should be organized in YOLO format:

• Images in dataset/train/images/ and dataset/val/images/
• Labels in dataset/train/labels/ and dataset/val/labels/
• Each label file should contain bounding box annotations in YOLO format

Training the Model

1. Prepare your dataset (if using custom data):

   python prepare_dataset.py

2. Start training:

   python train_new_model.py

3. Monitor training progress:

   • Check the runs/detect/ directory for training logs
   • View training plots and metrics
   • The best model will be saved as runs/detect/waste_classification_new/weights/best.pt

Training Parameters

Key training parameters (configurable in train_new_model.py):

• Epochs: 50 (default)
• Batch Size: 16 (GPU) / 8 (CPU)
• Image Size: 640x640
• Learning Rate: 0.01
• Device: CUDA (if available) / CPU

🔍 Inference

Command Line Inference

# Basic inference
python test_inference.py --model "path/to/model.pt" --image "path/to/image.jpg"

# With custom confidence threshold
python test_inference.py --model "path/to/model.pt" --image "path/to/image.jpg" --confidence 0.3

# Without saving result image
python test_inference.py --model "path/to/model.pt" --image "path/to/image.jpg" --no-save

Python API

from ultralytics import YOLO
from category_mapper import WasteCategoryMapper

# Load model
model = YOLO("path/to/model.pt")

# Load category mapper
mapper = WasteCategoryMapper()

# Run inference
results = model.predict("image.jpg", conf=0.25)

# Map to desired categories
mapped_results = mapper.map_prediction(model, results[0])

# Process results
for item in mapped_results:
    print(f"Category: {item['mapped_class']}, Confidence: {item['confidence']:.3f}")

🌐 Web Application

Launching the App

cd "streamlit-detection-tracking - app"
streamlit run app.py

The app will be available at http://localhost:8501

Features

• Image Upload: Upload images for waste classification
• Webcam Support: Real-time classification using your webcam
• Confidence Control: Adjust detection sensitivity
• Category Mapping: See how model categories map to your desired categories
• Detailed Results: View category counts and confidence scores

App Interface

1. Main Page: Shows supported waste categories
2. Sidebar: Model configuration and input source selection
3. Results Panel: Detailed detection results with category mapping
4. Mapping Info: Explanation of category conversions

🔧 Category Management

Viewing Categories

python category_manager.py

Testing Category Mapping

python category_mapper.py

Updating Categories

If you need to modify categories:

1. Edit data.yaml:

   names:
   - Glass
   - Organic
   - Others
   - Packaged
   - Plastic
   nc: 5

2. Retrain the model (if needed):

   python train_new_model.py

🐛 Troubleshooting

Common Issues

1. Model not found:

   • Check that best.pt exists in the weights directory
   • Verify the file path in settings.py

2. CUDA out of memory:

   • Reduce batch size in training parameters
   • Use CPU instead of GPU

3. Category mismatch:

   • Run python verify_model_categories.py to check categories
   • Use the category mapping solution

4. Poor detection accuracy:

   • Adjust confidence threshold
   • Ensure good image quality
   • Consider retraining with more data

Getting Help

• Check the console output for error messages
• Verify all dependencies are installed
• Ensure image paths are correct
• Review the category mapping if results seem incorrect

📊 Performance

Model Performance

• mAP50: ~0.85 (on validation set)
• Inference Speed: ~50ms per image (GPU)
• Model Size: ~50MB
• Input Size: 640x640 pixels

System Requirements

• Minimum: 4GB RAM, CPU-only
• Recommended: 8GB RAM, CUDA-compatible GPU
• Optimal: 16GB RAM, RTX 3060 or better

🔄 Future Improvements

• Support for more waste categories
• Mobile app development
• Real-time video processing
• Integration with IoT devices
• Multi-language support
• Batch processing capabilities

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙏 Acknowledgments

• Ultralytics for YOLOv8
• Streamlit for the web interface
• OpenCV for image processing
• The waste classification research community

---

Ready to classify waste? Start with the Quick Start guide above! 🚀