Published February 25, 2026 | Version 1.0.0

Synthetic Photogrammetric Dataset for Two-Media 3D Reconstruction: Shipwreck & Terrain

  • 1. Unit of Geometry and Surveying, Faculty of Engineering Sciences, University of Innsbruck
  • 2. Department of Geodesy and Geoinformation, TU Wien
  • 3. Institute of Photogrammetry and Remote Sensing, Karlsruhe Institute of Technology

Description

The data in this repository serves as a benchmark for the development of new image-based processing pipelines that consider two separate optical media, i.e., water and air, including refraction effects. Ground truth datasets representing the true geometry form the basis of the simulation.

The synthetic scene showcases an environment constructed from both real-world laser scanning data (Jamtal glacial valley - WGS84: 46.90° N, 10.17° E) and a synthetic CAD model of a sailing ship. Although the Jamtal scan depicts a riverbed, the original data were not acquired underwater and is only synthetically flooded. The ship was chosen because of the fine details of the ropes, which are well-suited for determining the quality of the reconstruction, and larger structured areas on the hull, where classic photogrammetric methods should perform well. In the simulation dataset the entire ship is submerged and large parts of the scene are submerged (up to 8 m depth), therefore the influence of refraction must be taken into account for an accurate reconstruction of these areas. The planar water surface serves as an idealized baseline to evaluate geometric refraction correction without the interference of dynamic wave patterns.

Further technical details can be found in description_dataset.pdf.

Files

description_dataset.pdf

Files (499.9 MB)

Name Size
md5:e24b46609e328914f5288cb4ed75e52f
499.5 MB Preview Download
md5:ac6cb254472b56a25234d0d4291317ce
437.2 kB Preview Download
md5:d3a8ecaaa935284fe990ce5a5f87bb0b
1.8 kB Preview Download

Additional details

Funding

FWF Austrian Science Fund
Bathymetrische 3D-Rekonstruktion mit Neural Radiance Fields 10.55776/PIN1353223
Deutsche Forschungsgemeinschaft
Bathymetrische 3D-Rekonstruktion mit Neural Radiance Fields 538522540

References

  • Schulte, F., Brezovsky, M., Günthner, A., Jutzi, B., Mandlburger, G., & Winiwarter, L. (2025). Simulation and validation of underwater scenes for two-media optical 3D reconstruction. The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XLVIII-2/W10-2025, 271–278. https://doi.org/10.5194/isprs-archives-xlviii-2-w10-2025-271-2025