Cumulative Assessment for Urban 3D Modeling

TL;DR

This paper introduces a cumulative assessment metric for urban 3D modeling that evaluates errors from semantic segmentation, stereo reconstruction, and model fitting, supported by public datasets and open-source tools.

Contribution

It proposes a comprehensive error metric for urban 3D modeling and provides an end-to-end baseline with datasets and open-source extensions.

Findings

New cumulative error metric for urban 3D modeling

Public datasets and open-source baseline provided

Facilitates standardized evaluation and further research

Abstract

Urban 3D modeling from satellite images requires accurate semantic segmentation to delineate urban features, multiple view stereo for 3D reconstruction of surface heights, and 3D model fitting to produce compact models with accurate surface slopes. In this work, we present a cumulative assessment metric that succinctly captures error contributions from each of these components. We demonstrate our approach by providing challenging public datasets and extending two open source projects to provide an end-to-end 3D modeling baseline solution to stimulate further research and evaluation with a public leaderboard.