Computer Vision Waste Sorting Assistant

📚 About The Project

The Computer Vision-Enhanced Waste Sorting Assistant is a machine learning-powered system designed to enhance waste classification accuracy and improve recycling habits through real-time waste detection and data visualization. By utilizing YOLOv11 and a user-friendly interface, the project combines state-of-the-art object detection with actionable data insights to empower users and municipalities in their waste management efforts.

👥 Team

• Xinyi Wang - GitHub Profile
• Yijie Cao - GitHub Profile
• Min Ren GitHub Profile

🎯 Project Structure

CS5330FINAL/
├── .idea/                    # IDE configuration files
├── csv/                      # Waste management data (CSV files for dashboard analysis)
├── flagged/                  # Placeholder for flagged items (optional usage)
├── runs/                     # YOLOv11 training results and logs
├── ultralytics/              # YOLOv11 implementation and configurations
├── venv/                     # Virtual environment for project dependencies
├── .gitattributes            # Git LFS tracking information for large files
├── bar_chart.png             # Sample visualization of waste trends
├── best.pt                   # YOLOv11 model weights (best performance)
├── Data analysis.py          # Dashboard implementation code
├── end.py                    # Post-processing script for YOLO detections
├── Image_stitching.py        # Script for creating multi-object collages for training
├── Interface system.py       # Gradio-based interface for waste detection
├── mytrain.py                # Custom YOLOv5 training script
├── output_collage.jpg        # Sample collage image for training
├── output_collage2.jpg       # Sample collage image for training
├── output_collage3.jpg       # Sample collage image for training
├── output_collage4.jpg       # Sample collage image for training
├── output_collage5.jpg       # Sample collage image for training
├── predict.py                # Script for inference on test images
├── README.md                 # Project overview
├── requirements.txt          # List of dependencies for the project
├── results.csv               # Training results and metrics
├── rubbish.yml               # YOLOv11 dataset configuration file
├── test.py                   # Script for evaluating model performance
├── test.txt                  # Placeholder for testing data
├── txt_csv.py                # Script for converting YOLO text annotations to CSV format
├── visualization.png         # Dashboard visualization example
├── yolo11m.pt                # YOLOv11 model weights (latest)

📊 Documentation

This project includes comprehensive documentation to help users understand, implement, and utilize the Computer Vision-Enhanced Waste Sorting Assistant.

Key Sections:

1. System Overview
   Detailed explanation of the waste sorting assistant, its architecture, and its functionalities.
   Read the system overview here.

2. Dataset Preparation
   Step-by-step instructions on how the dataset was prepared, including preprocessing and augmentation techniques.
   Dataset Preparation Documentation).

3. Model Training and Configuration
   Details on the YOLOv5-based model, including architecture, parameters, training workflow, and hyperparameter tuning.
   Model Training Guide).

4. Dashboard Integration
   Documentation for the Gradio-based dashboard, including its features, configuration, and deployment process.
   Dashboard Guide).

5. Results and Analysis
   Insights and analysis of model performance, evaluation metrics, and waste trends from the dashboard.
   Results Analysis.

💡 Key Features

• Real-Time Waste Detection

  • Identifies multiple waste categories in a single image: recyclable, hazardous, kitchen, and other waste.
  • Provides bounding box visualizations and tailored recycling recommendations.
  • Powered by YOLOv11, offering high-speed and reliable object detection.

• Interactive Data Analysis Dashboard

  • Upload CSV files for waste data analysis.
  • Generate stacked bar charts, line graphs, and pie charts.
  • Explore trends such as monthly tonnage, classification accuracy, and recycling rates.

• User-Friendly Interfaces

  • Built using Gradio for accessibility and interactivity.
  • Huggingface Link: https://huggingface.co/spaces/96philly/5330final
  • Supports exporting of visualizations and reports for further analysis.

🛠️ How It Works

This section provides a high-level explanation of how the Computer Vision-Enhanced Waste Sorting Assistant operates:

1. Dataset Preparation

   • A dataset of 2,744 waste images was collected and categorized into four classes:
     Recyclable, Hazardous, Kitchen, and Other waste.
   • Images were annotated in YOLO-compatible format, and augmentations like horizontal flipping and mosaic augmentation were applied to improve the model's robustness.

2. Model Training

   • The YOLOv5 model was fine-tuned for waste detection using the prepared dataset.
   • Key parameters:
     • Learning rate: 0.01
     • Batch size: 32
     • Epochs: 400
   • Loss functions included localization, classification, and objectness loss.
   • Metrics such as mAP@50 were used for evaluation.

3. Gradio Dashboard

   • A Gradio-based interactive dashboard allows users to:
     • Upload CSV files for waste trend analysis.
     • Visualize monthly tonnage, classification accuracy, and recycling rates using line charts, bar charts, and pie charts.

4. Real-Time Waste Classification

   • The model predicts bounding boxes for waste objects in uploaded images.
   • Results include waste categories, locations, and recycling suggestions, enabling users to make data-driven decisions.

---

📊 Model Performance

The model demonstrated high detection accuracy and efficiency across various scenarios:

1. Performance Metrics

   • Mean Average Precision (mAP@50): 83.49%
   • mAP@50-95: 74.97%
   • Precision: 83.49%
   • Recall: 82.28%
   • Losses:
     • Box Loss: 0.66185
     • Classification Loss: 0.39625

2. Detection Speed

   • The model completed 400 epochs in approximately 3 hours, achieving near real-time predictions.

3. Evaluation Insights

   • Confusion Matrix Analysis: Showed strong classification for "Recyclable" and "Kitchen Waste" categories but identified opportunities for improvement in "Other" and "Hazardous Waste" detection.
   • Loss Convergence: Training and validation loss curves indicate stable convergence across all loss components (see results.png).

4. Visualization

   • The Gradio dashboard offers detailed performance tracking and interactive waste analysis tools.
   • Example charts （burnaby) include classification accuracy trends, recycling rates, and monthly waste tonnage distribution.

   

   

   

📝 Course Information

This final project was developed as part of the Computer Vision course at Northeastern University.

🔎 References

1. YOLOv5 Documentation
   Official YOLOv5 documentation and resources.
   Read more here.

2. Microsoft Research Asia
   For insights into advanced computer vision models and research on transformers.
   Visit here.

3. Gradio Documentation
   For building interactive machine learning interfaces.
   Learn more.

4. Metro Vancouver Waste Management Summary 2021
   Source of municipal waste management data for analysis and dashboard integration.
   Access the report here.

5. PyTorch Framework
   Deep learning framework used for implementing YOLOv5 and custom scripts.
   Explore PyTorch.

6. RoboFlow Dataset
   Trash detection dataset referenced for categorization and augmentation techniques.
   Visit RoboFlow.

📖 Citation

If you find this project helpful, please consider citing:

@project{waste_sorting_assistant2024,
    title={Computer Vision-Enhanced Waste Sorting Assistant},
    author={Xinyi Wang},
    institution={Northeastern University},
    year={2024},
    note={Deep Learning Course Project}
}

For academic use, please ensure to cite both the original work and our educational materials appropriately. }

## 🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

1. Fork the Project
2. Create your Feature Branch (`git checkout -b feature/AmazingFeature`)
3. Commit your Changes (`git commit -m 'Add some AmazingFeature'`)
4. Push to the Branch (`git push origin feature/AmazingFeature`)
5. Open a Pull Request

## 📧 Contact

For questions and feedback, please open an issue in this repository or contact team members directly.

## 🙏 Acknowledgments

- Ultralytics for the YOLOv11 framework.
-Gradio for the user interface library.
-Northeastern University for course guidance and support.

---

Made with ❤️ by Our Team
</div>