🗑️ Object Detection for Waste Management Project

📘 Overview

This repository contains the implementation of a multi-class waste detection model built using the YOLOv8s architecture. The project aims to automate the identification and classification of various types of waste to assist in efficient waste management and recycling systems. The model was trained on the TACO (Trash Annotations in Context) dataset, which includes 18 waste categories such as plastic bottles, cans, paper, and glass.

---

🎯 Project Objective

The goal of this project is to:

• Detect and classify different types of waste items in real-world images.
• Assist in smart waste management using computer vision and deep learning.
• Contribute toward environmental sustainability by enabling automated waste sorting.

---

🧠 Model Architecture

• Base Model: YOLOv8s (You Only Look Once, version 8 - Small)
• Framework: Ultralytics YOLOv8 (PyTorch backend)
• Number of Classes: 18 (based on the TACO dataset)
• Image Size: 640 × 640 pixels
• Epochs: 100
• Optimizer: Adam
• Loss Function: Object Detection Loss (composed of box loss, cls loss, and DFL loss)

---

📊 Dataset Description

Dataset Name: TACO Trash Detection Dataset
Source: Public dataset from Kaggle
Description: A dataset of images annotated with 18 waste categories for object detection tasks.
Each image contains one or more waste items labeled under:

Plastic, Metal, Paper, Glass, Cardboard, Textile, Organic, etc.

---

⚙️ Training Environment

• Platform: Google Colab
• Hardware: GPU (Tesla T4)
• Frameworks & Libraries:
  • Python 3.10
  • PyTorch
  • Ultralytics YOLOv8
  • OpenCV
  • NumPy
  • Matplotlib

---

📈 Results and Evaluation

Metric	Value
mAP@0.5	0.67
mAP@0.5:0.95	0.45
Precision	0.72
Recall	0.68
Validation Loss	1.041
Validation Accuracy	0.44

The model shows strong generalization on diverse waste items and performs well in real-world scenarios.

---

🧾 Notebook File

The main notebook in this repository is: taco-trash-detection-yolov8s.ipynb

It includes:

• Dataset preparation (TACO)
• Model training (YOLOv8s)
• Evaluation and visualization of predictions
• Detection results on validation images

---

🚀 How to Run the Project

1. Clone this repository:

   git clone https://github.com/<your-username>/object-detection-for-waste-management-project.git
   cd object-detection-for-waste-management-project

2. Install dependencies:

   pip install ultralytics opencv-python matplotlib numpy

3. Open and run the notebook:

   jupyter notebook taco-trash-detection-yolov8s.ipynb

4. The model will train and evaluate automatically using the configured parameters.

---

📂 Repository Structure

object-detection-for-waste-management-project/
│
├── taco-trash-detection-yolov8s.ipynb      # Main training and evaluation notebook
├── README.md                               # Project documentation
└── data/                                   # Dataset (optional local directory)

---

📚 Future Work

• Experiment with larger YOLOv8 variants (YOLOv8m or YOLOv8l)
• Deploy as a real-time waste detection system using a webcam
• Integrate with IoT devices for smart bin automation

---

🧑‍💻 Author

Assad Ullah Khan
Research Assistant – Digital Image Processing Lab
Islamia College University, Peshawar
📧 Email: assadullahkhan556@gmail.com

---

🏆 Acknowledgment

Special thanks to Dr. Mohammad Sajjad (Associate Professor) for his supervision and guidance throughout this research project.

---

📜 License

This project is licensed under the MIT License – feel free to use and modify for research or educational purposes.

---

“Automating waste detection through AI and vision systems can help pave the way for a cleaner, smarter, and more sustainable planet.”

---

Made by Assad Ullah Khan
Research Assistant at DIP lab Islamia College Peshawar