Abstract
TL;DR: Aerial to street-level localization dataset covering both urban, suburban and rural areas. Contains street-level images with accurate 6 DoF poses and associated aerial images images from Denmark.
Place recognition and visual localization are particularly challenging in wide baseline configurations. In this paper, we present the Danish Airs and Grounds (DAG) dataset, a large collection of street-level and aerial images targeting such cases. Its main challenge lies in the extreme viewing-angle difference between query and reference images with consequent changes in illumination and perspective. We propose a map-to-image re-localization pipeline that first estimates a dense 3D reconstruction from the aerial images and then matches query street-level images to street-level renderings of the 3D model.
Aerial images used to construct the dataset
Method Overview
Street-level to aerial localization pipeline. We generate a dense and accurate multi-view 3D reconstruction from aerial images. Using this 3D model we can render a database of photo realistic images for real-image-to-render retrieval from actual street-view queries. Our experiments show that this is an effective pipeline for air-to-ground retrieval and localization.