Building Detector

An AI-powered web application for detecting and extracting building footprints from satellite imagery

Overview

Building Detector is a sophisticated tool that leverages deep learning to identify and extract building geometries from satellite imagery. The application provides an intuitive web interface where users can:

✅ Select areas of interest on an interactive map
✅ Download high-resolution satellite imagery
✅ Place reference points on specific buildings to guide the detection algorithm
✅ Process images to identify marked buildings
✅ Export results in GeoJSON format or OpenStreetMap-compatible format

Architecture

The system uses a two-tier architecture:

• Frontend (Flask Web App): Provides the user interface and handles satellite imagery downloads
• Backend (Colab ML Server): Runs the machine learning model using Segment Anything Model 2 (SAM2)

Key Features

• 🗺️ Interactive map-based interface for area selection
• 📡 One-click satellite image download for any region
• 🎯 Point-based detection control - users specify exactly which buildings to detect
• 🔧 Building regularization for clean, precise geometries
• 📊 Dual export options:
  • Standard GeoJSON with building metadata
  • OpenStreetMap-compatible GeoJSON for easy OSM contributions

Project Structure

building-detector/
├── main.py                    # Main entry point
├── requirements.txt           # Python dependencies
├── .gitignore                # Git ignore rules
├── LICENSE                   # MIT License
│
├── app/                      # Main application directory
│   ├── app.py               # Flask server & routing logic
│   ├── config.py            # Configuration settings
│   ├── config.py.template   # Configuration template
│   ├── static/              # Static assets (CSS, JS, images)
│   ├── templates/           # HTML templates
│   └── utils/               # Utility functions
│
├── local_backend/           # 🆕 Local ML backend (SAM2)
│   ├── local_server.py     # Local SAM2 server
│   ├── requirements.txt    # ML dependencies
│   └── README.md           # Setup instructions
│
├── uploads/                 # Temporary storage (auto-created)
├── UML/                    # System documentation
├── building-detector.ipynb # Colab notebook
└── .env.example           # Environment template

Technology Stack

• Backend: Flask, Python 3.8+
• Frontend: HTML5, CSS3, JavaScript, Leaflet.js
• AI/ML: Segment Anything Model 2 (SAM2)
• Geospatial: leafmap, GeoJSON
• Deployment: ngrok (for Colab connectivity)

Prerequisites

• Python 3.8 or higher
• Modern web browser with JavaScript enabled
• Internet connection for satellite imagery and ML processing
• Google Colab account (for ML backend)

Quick Start

1. Clone the Repository

git clone https://github.com/yourusername/building-detector.git
cd building-detector

2. Set Up Virtual Environment

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

3. Install Dependencies

pip install -r requirements.txt

4. Run the Application

python main.py

5. Access the Web Interface

Open your browser and navigate to: http://127.0.0.1:5000/

Configuration

Option 1: Google Colab Backend (Recommended)

1. Access the Colab notebook: Building Detector ML Server
2. Make a copy to your Google Drive
3. Get ngrok token: Register at ngrok.com and copy your auth token
4. Run the notebook: Enter your ngrok token when prompted
5. Update configuration: Copy the ngrok URL to app/config.py:

   COLAB_SERVER_URL = 'https://your-ngrok-url.ngrok.io'

Option 2: Local Backend (Full SAM2 Implementation)

NEW! Run the complete AI pipeline locally using the same SAM2 model as Colab:

1. Install local backend:

   cd local_backend
   pip install -r requirements.txt

2. Start local ML server:

   python local_server.py  # Runs on port 5001

3. Configure main app:

   export COLAB_SERVER_URL='http://127.0.0.1:5001'
   python main.py

Benefits: No cloud dependencies, data privacy, no session timeouts Requirements: 8GB+ RAM, Python 3.8+, optional GPU for speed

See local_backend/README.md for detailed setup instructions.

Usage Guide

1. 📍 Select Area: Navigate to your area of interest on the interactive map
2. 📡 Download Imagery: Click "Download Satellite Image" for the visible region
3. 🎯 Place Points: Click on buildings you want to detect (each click marks a building)
4. ⚙️ Process: Click "Detect Buildings" to start the AI processing
5. 💾 Export Results: Download detected buildings as GeoJSON files

Security & Configuration

• API keys and credentials are externalized to environment variables
• Sensitive data is excluded from version control via .gitignore
• Upload directory is automatically cleaned between sessions

Contributing

Contributions are welcome! Please feel free to submit issues and enhancement requests.

License

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

Author

Silas Pignotti

---