VRAK3D - Virtual archaeology, when knowledge meets 3D

The VRAK3D project lies at the crossroads of several disciplines: archaeology, architecture, knowledge representation, and is based on innovative technologies: computer vision, deep learning, virtual and augmented reality. Based on the observation that, while obtaining 3D models in the field of heritage is now virtually resolved and no longer part of the field of Research, the link between knowledge and geometry remains embryonic.

This project is divided into four phases around the Glanum site, which has been extensively excavated since the 1920s:

1. Development of a 3D model of the site capable of supporting the archaeological knowledge already acquired, as well as serving as a user interface to help generate new knowledge (drone and terrestrial photogrammetry).
2. Development and formalization of a corpus of archaeological objects using ontologies aligned with existing work in the community (CRM-CIDOC, for example). This formalization of knowledge will describe both morphological aspects and the relationships that these elements may have with each other. For example, the topological relationship between a capital and an architrave.
3. Research these objects using deep learning on all the photos taken to obtain the 3D model. The search for these elements includes a phase of labeling the objects on the photos (e.g. blocks, capitals, column shafts, architraves, etc.), then searching them on all the photos of the site, and finally a 3D localization using the orientations of the photos calculated by photogrammetry.
4. The virtual reality aspects of the project will be broken down into several faculties: virtual tour of the site (headset and web), interface with current knowledge (e.g. corpus, archives, bibliography), manipulation of corpus elements identified on the 3D model by satisfying relationship constraints defined in ontologies, partial and virtual anastylosis with geometric quantification of relationship relevance.