TerraSky3D: Multi-View Reconstructions of European Landmarks in 4K

2026-03-30

AI Summary

Existing 3D reconstruction datasets often suffer from low resolution, limited scene scale, restricted viewpoints, or inconsistent image quality, hindering the training and evaluation of advanced algorithms. To address these limitations, this work introduces TerraSky3D—a high-resolution, large-scale multi-view 3D reconstruction benchmark comprising 150 European landmark scenes. The dataset integrates ground-level and aerial drone imagery, providing 50,000 4K images accompanied by accurate camera poses, calibration parameters, and depth maps. TerraSky3D is the first to systematically combine high-quality heterogeneous viewpoints under a unified acquisition protocol and comprehensive geometric annotations, establishing a challenging yet standardized benchmark for 3D vision research.

Abstract

Despite the growing need for data of more and more sophisticated 3D reconstruction pipelines, we can still observe a scarcity of suitable public datasets. Existing 3D datasets are either low resolution, limited to a small amount of scenes, based on images of varying quality because retrieved from the internet, or limited to specific capturing scenarios. Motivated by this lack of suitable 3D datasets, we captured TerraSky3D, a high-resolution large-scale 3D reconstruction dataset comprising 50,000 images divided into 150 ground, aerial, and mixed scenes. The dataset focuses on European landmarks and comes with curated calibration data, camera poses, and depth maps. TerraSky3D tries to answer the need for challenging dataset that can be used to train and evaluate 3D reconstruction-related pipelines.

Problem

• 3D reconstruction
• dataset scarcity
• high-resolution data
• multi-view imagery
• camera calibration

Innovation

• 3D reconstruction
• multi-view dataset
• high-resolution imaging
• camera pose estimation
• depth maps