National XR Day 2025

Gaussian Splatting in XR: Photoreal 6DoF
2025-07-02 , Theil C1-6 (60p)

Since its initial release in 2023, gaussian splatting has fundamentally altered the way we can create photoreal 3D scenes for XR experiences. This workshop discusses its applications in an educational context. The evolving technical aspects of splat creation and workflows will be covered in depth, followed by a live demonstration that showcases an ongoing criminology project at Leiden University.


3D Gaussian Splatting (3DGS), first released by Inria in 2023, is a novel method of constructing 3D scenes from simple photos or videos. Belonging to the broader category of radiance fields, 3DGS creates photoreal environments that can be quickly rendered in real time, marking a departure from previous methods like NeRF.

Across 3D applications, this means that scenes or objects may be quickly constructed, providing a new means of rapid prototyping. VR creators have long had to contend with lengthy 3D modeling times in order to achieve results that approach photorealism, or compromise with lower fidelity techniques such as photogrammetry. 3DGS aims to solve these problems with low manhours and highly realistic renderings, making it an attractive solution for XR.

Nonetheless, the field is quite new and not without a number of stumbling blocks. These can largely be solved through 3DGS-optimized workflows and (of course) by adding significantly more compute capabilities with the use of virtual machines. This workshop covers these aspects in depth, and culminates in a live demonstration of a strong potential use case: urban planning interventions.

Together with the City of Leiden and the Leiden Law School, the Leiden Learning and Innovation Centre has created a prototype that explores urban safety research in criminology by simulating various interventions using 3DGS scenes of municipal locations. These simulations allow planners and policymakers to “pre-visualize” proposed interventions and test them with research participants, providing a model to evaluate efficacy prior to physical implementation.