EcoGuard AI 🌍🛡️

EcoGuard AI is a comprehensive, AI-powered computer vision platform designed to tackle industrial pollution, enforce worker safety compliance, and monitor environmental health.

Developed at ESPRIT (École Supérieure Privée d'Ingénierie et de Technologies / Esprit School of Engineering), EcoGuard utilizes advanced deep learning architectures (YOLOv8 & PyTorch U-Net) and is deployed via a modern Django web application. It provides real-time, interactive insights into high-risk industrial zones (like mining operations) and their surrounding ecosystems.

---

✨ Key Features

👷‍♂️ Industrial Worker Safety

• PPE Compliance Tracking: Real-time YOLOv8 models fine-tuned to detect hard hats, medical masks, and heavy-duty gas masks.
• Explainable AI (XAI): EigenCAM integrations to visualize exactly what the model is looking at when making safety predictions.

🏭 Environmental & Mining Monitoring

• Mining Area Segmentation: PyTorch U-Net models trained on satellite imagery to precisely calculate the footprint of legal and illegal mining operations.
• Soil Health Analysis: The custom "Ghada" U-Net model evaluates land degradation and soil health based on the proximity and density of mining excavations.
• Deforestation Tracking: Automated tracking of forest loss around industrial zones using custom satellite segmentation.
• Smoke & Fire Detection: Early warning systems for industrial exhaust and wildfires.

🐟 Wildlife & Ecosystem Tracking

• Aquatic Contamination Analysis: Advanced fish behavior tracking. The system uses Re-Identification (Re-ID) and trajectory mapping to classify fish swimming patterns (Normal vs. Stressed) as an early indicator of phosphogypsum water contamination.
• Terrestrial Wildlife: YOLO-based animal detection to monitor wildlife displacement near mining zones.

🔐 Smart Authentication & Administration

• OCR-Powered Registration: Employees register by uploading their ID cards. The system uses EasyOCR to automatically extract their First and Last names, auto-generate a corporate email (first_last@EcoGuard.ai), and generate a highly secure random password.
• Approval Workflow: New accounts are placed in a "Pending Approval" state.
• Admin Dashboard: Superusers have access to a User Management interface to view ID cards and securely Approve, Reject, or Delete employee access.

---

🛠️ Technology Stack

Machine Learning & Computer Vision:

• PyTorch / Torchvision
• Ultralytics YOLOv8
• Segmentation Models PyTorch (SMP)
• OpenCV / NumPy
• EasyOCR

Web Development:

• Django (Python Web Framework)
• SQLite3 (Database)
• Bootstrap 5 (Mazer Admin Template)
• Chart.js / ApexCharts (Data Visualization)

---

🚀 Installation & Setup

1. Clone the repository:

   git clone https://github.com/zaabola/Projet_Mining_Pollution.git
   cd Projet_Mining_Pollution

2. Create and activate a virtual environment:

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

3. Install dependencies: (Ensure you have PyTorch installed according to your CUDA version first)

   pip install -r requirements.txt

4. Run Database Migrations:

   cd web_app
   python manage.py makemigrations
   python manage.py migrate

5. Create an Admin User:

   python manage.py createsuperuser

6. Start the Development Server:

   python manage.py runserver

   Access the dashboard at http://localhost:8000.

---

📸 Screenshots

---

📄 Credits & Academic Context

This project was developed at ESPRIT (Esprit School of Engineering) as part of an advanced environmental technology and artificial intelligence initiative. All rights reserved.