Overview contents
Research question
How far can the time and manual work needed to digitise a real place be reduced while maintaining visual quality and usability suitable for commercial use?
Technical context
3D Gaussian Splatting represents a scene through numerous coloured, semi-transparent three-dimensional elements oriented in space. These elements are optimised by comparing images of the same environment captured from different viewpoints.
Compared with a reconstruction modelled entirely by hand, the starting point is the real space as captured. The work does not disappear: it shifts to the capture protocol, processing, data optimisation and interactive experience design.
What we are studying
- 01
Repeatable capture protocols using smartphones, cameras and drones.
- 02
Reconstruction, cleanup and optimisation of scenes for different devices.
- 03
Delivery through a browser or headset and progressive handling of large datasets.
- 04
Adding hotspots, information and editable 3D components.
Applications to explore
- Showrooms, trade fair stands and commercial spaces that can be visited remotely.
- Tours of properties, hotels, wineries, museums and cultural heritage sites.
- Visual documentation of construction sites, facilities and production processes.
- Experiences for vineyards, agrifood supply chains and contextual training.
How it can combine with other technologies
A 3D reconstruction can provide the visual layer of a wider system. Computer vision can recognise elements in the scene, and a Visual Positioning System can estimate position and orientation. In a configurator, the captured environment can also coexist with products represented by interchangeable 3D meshes.
What a prototype needs to establish
- V01
The quality achievable with different devices and capture conditions.
- V02
Dataset size, loading times and smooth performance in target browsers and headsets.
- V03
The working time and skills needed to capture and update an environment.
- V04
The actual usefulness of hotspots and interactive layers for the specific use case.
Known limitations and risks
Glass, mirrors, uniform surfaces, movement and changes in lighting can produce artefacts.
A visually convincing representation does not automatically provide metric measurements, collisions, separable objects or parametric geometry.
Technical references