This repository addresses the algorithmic challenges of the IARPA CORE3D program. The goal of this software is to reconstruct semantically meaningful 3D models of buildings and other man-made structures from satellite imagery.

This repository contains the algorithms to solve the CORE3D problem, but the user interface and cloud-based processing infrastructure are provided in a separate project called Resonant Geo. The algorithms in this repository are written in Python or at least provide a Python interface.

Clone this repository with its submodules by running:

git clone --recursive [email protected]:Kitware/Danesfield.git

To fetch the latest version of this repository and its submodules, run:

git pull
git submodule update --init --recursive

The first step in running or developing Danesfield code is to obtain the correct development environment. The Danesfield algorithms require a number of dependencies on geospatial and computer vision libraries. Provided with this repository are instructions for configuring a development environment with Conda. Conda provides a consistent development environment with a known configuration of dependencies versions. Follow the directions in deployment/conda/README.rst to setup this environment.

This repository has also been built into a Docker image, which includes the required conda environment. The image was built using the Dockerfile included in this repository. As some of the Danesfield algorithms require a GPU, you'll need to have NVIDIA Docker installed, and use the nvidia-docker command when running the image.

The Danesfield project is organized as follows:

• danesfield This directory is where the danesfield algorithmic modules live.
• tools This directory contains command line tools to execute the Danesfield algorithms.

GDAL/OGR cookbook

Workshop: Raster and vector processing with GDAL