🛢️ AI SpillGuard Pro

Enterprise-Grade Oil Spill Detection System using Deep Learning & Satellite Imagery

AI SpillGuard Pro is a production-ready web application that leverages U-Net semantic segmentation with ResNet34 backbone to detect and classify oil spills in SAR (Synthetic Aperture Radar) satellite imagery. The system provides real-time detection, intelligent alerting, persistent storage, and comprehensive analytics for marine environmental monitoring.

✨ Key Features

• 🎯 Multi-Class Segmentation: Distinguishes between oil spills, look-alikes, ships/wakes, and background with pixel-level precision.
• 🚨 Intelligent Alert System: Configurable threshold-based monitoring with priority-ranked critical alerts.
• 📊 Real-Time Visualization: Interactive overlay rendering with adjustable transparency and class toggles.
• 💾 Persistent Storage: Automatic saving of detection results, images, and searchable history.
• 📈 Analytics Dashboard: Comprehensive statistics, coverage distribution, and historical trend analysis.
• 🔌 REST API Integration: FastAPI-based programmatic access with Docker deployment support.
• ⚡ GPU Acceleration: Optimized PyTorch inference with CUDA support for high-throughput processing.
• 🎨 Professional UI: Streamlit-based interface with responsive design and enterprise styling.

---

🛠️ Tech Stack

Component	Technology
Deep Learning Framework	PyTorch 2.x
Model Architecture	U-Net with ResNet34 encoder (segmentation_models_pytorch)
Frontend	Streamlit 1.x
Backend API	FastAPI (optional deployment)
Image Processing	OpenCV, PIL, NumPy
Data Storage	JSON-based history tracking
Visualization	Matplotlib, Plotly
Development	Python 3.8+, Jupyter Notebooks

---

🏗️ Architecture Overview

┌─────────────────────────────────────────────────────────────┐
│                      Streamlit Frontend                     │
│  (User Interface + Visualization + Controls)                │
└──────────────────────┬──────────────────────────────────────┘
                       │
          ┌─────────────┼─────────────┐
          ▼             ▼             ▼
┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│OilSpillDetector│ │AlertSystem │ │StorageManager│
│(Inference)   │ │(Monitoring) │ │(Persistence) │
└──────┬──────┘ └─────────────┘ └─────────────┘
       │
       ▼
┌─────────────────────┐
│   U-Net Model       │
│   (best_model.pth)  │
│   ResNet34 Encoder  │
└─────────────────────┘
       │
       ▼
┌─────────────────────┐
│ Detection Results   │
│ • Segmentation Mask │
│ • Statistics        │
│ • Alerts            │
└─────────────────────┘

Component Interactions:

• OilSpillDetector: Handles model loading, preprocessing, inference, and post-processing.
• AlertSystem: Monitors detection statistics against configurable thresholds.
• StorageManager: Persists detection history to history.json.
• Streamlit UI: Orchestrates user interactions across detection, history, and API tabs.

---

📦 Installation & Setup

Prerequisites

• Python 3.8 or higher
• CUDA-compatible GPU (optional, for acceleration)
• 4GB+ RAM recommended

Step 1: Clone Repository

git clone <repository-url>
cd <project-directory>

Step 2: Install Dependencies

pip install -r requirements.txt

Step 3: Verify Model Files

Ensure best_model.pth exists in the project root:

ls -lh best_model.pth

Step 4: Prepare Dataset (Optional - Training Only)

If retraining the model:

python src/preprocess_train.py
python src/preprocess_val.py
python src/preprocess_test.py

Step 5: Launch Application

streamlit run app.py

Access the application at http://localhost:8501

---

🚀 Usage Examples

Web Interface (Streamlit)

1. Upload Image for Detection: Navigate to "Detection" tab, click "Browse files", and view results.
2. Configure Alert Thresholds: Sidebar → Alert Thresholds (Adjust sliders for Oil Spill, Look-alike, etc.).
3. Export Detection History: Navigate to "History" tab and click "Export History (CSV)".

Python API (Programmatic)

from app import OilSpillDetector, AlertSystem
import cv2

# Initialize detector
detector = OilSpillDetector("best_model.pth")
alert_system = AlertSystem()

# Load image (ensure RGB format)
image = cv2.imread("satellite_image.png")
image_rgb = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)

# Run detection
results = detector.predict(image_rgb)

# Access results
print(f"Oil Spill Coverage: {results['statistics']['Oil Spill']['percentage']:.2f}%")

# Check alerts
alerts = alert_system.check_alerts(results["statistics"])
for alert in alerts:
    print(f"{alert['severity']}: {alert['message']}")

---

📡 API Reference

REST API Endpoints (FastAPI)

Base URL: http://localhost:8000/api/v1

• **POST /detect**: Detect oil spills in uploaded image.
• **GET /health**: Health check endpoint.

---

🎨 Class Definitions

Class ID	Class Name	RGB Color	Hex Code
0	Background	(0, 0, 0)	#000000
1	Oil Spill	(255, 0, 124)	#FF007C
2	Look-alike	(255, 204, 51)	#FFCC33
3	Ship/Wake	(51, 221, 255)	#33DDFF

---

🐳 Docker Deployment

Build and Run

# Build Image
docker build -t ai-spillguard-api .

# Run Container
docker run -p 8000:8000 ai-spillguard-api

---

🤝 Contribution Guidelines

A contribution is considered complete when:

• ✅ Follows PEP 8 style guidelines.
• ✅ Type hints and Docstrings are added.
• ✅ Unit tests pass without regression.
• ✅ GPU/CPU compatibility is maintained.

---

📝 License

This project is licensed under the MIT License.

Copyright (c) 2025 Prem Chand

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files.

Built with ❤️ for Environmental Protection and Marine Conservation.

---