Roadmap

The FHNW Muttenz visual localization dataset is an ongoing effort. This page tracks planned extensions and gives an idea of where the project is heading. Release dates depend on the review process of the associated publications, so we give status rather than target dates. This page is updated as items progress; last updated 28 July 2026.

Planned extensions

Additional query sequences (In processing)

The FHNW Muttenz visual localization dataset currently contains only a limited range of challenging conditions. Currently, we are working on extending our dataset by incorporating additional areas along the road network and by including more challenging environmental conditions, such as diverse weather scenarios, seasonal changes, and, in particular, imagery acquired during dusk and nighttime. Query sequences in five additional scenes using three different cameras were captured from March to May 2026. Georeferencing and co-registration are in progress.

Aerial imagery (In processing)

By providing precisely georeferenced and co-registered aerial imagery, the FHNW Muttenz visual localization dataset will also enable the validation of cross-view visual localization methods. Oblique imagery of the full road network was captured in April/May 2026 using a DJI Mini 3 Pro. Georeferencing and co-registration are in progress.

Reference mapping campaign (Scheduled)

A third reference mapping campaign is planned. This campaign will cover the complete study area and will take place in late summer / early autumn 2026. The imagery will be co-registered to the existing two reference mapping campaigns with a targeted accuracy in the millimeter range (better than 1 cm).

Release of raw image data (Planned)

Currently, all query images provided are undistorted, using the camera calibration parameters obtained during ground-truth pose generation. To enable real-world applications, visual localization methods need to work with uncalibrated imagery. Therefore, we plan to release the raw imagery.