4Sat2OSM: Satellite Building Detection

Overview

This project implements building detection from satellite imagery using the SpaceNet 2 dataset. The system uses deep learning models to identify and segment building footprints in satellite images, which can later be processed for OpenStreetMap integration.

Features

• Building Detection: Automatic detection of buildings from satellite imagery
• Deep Learning Model: Uses UNet++ architecture for accurate segmentation
• SpaceNet 2 Integration: Designed to work with SpaceNet 2 dataset
• Modular Design: Clean, understandable code structure
• Visualization Tools: Built-in tools for result visualization and analysis

Project Structure

4Sat2OSM/
├── data/                     # Dataset directory
│   └── spacenet2/           # SpaceNet 2 dataset
│       ├── images/          # Satellite images
│       └── masks/           # Building masks
├── models/                  # Trained model weights
├── results/                 # Output results and visualizations
├── notebooks/               # Jupyter notebooks
│   └── BuildingDetection.ipynb
├── src/                     # Source code
│   └── 4sat2osm/
└── README.md

Installation

This project uses uv as the Python package manager. Make sure you have uv installed:

# Install uv if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install project dependencies
uv sync