The Promises of Gaussian Splatting for Heritage Preservation
The emergence of three-dimensional representation technologies, particularly Neural Radiance Fields (NeRFs) and Gaussian Splatting (GS), opens new perspectives for the preservation and transmission of intangible cultural heritage. These advances offer unprecedented possibilities for capturing and reproducing technical gestures in the field of craftsmanship.
The true potential of these technologies lies in their ability to simultaneously capture several essential dimensions of cultural heritage. As demonstrated by Kerbl et al. [1] in their work on “3D Gaussian Splatting for Real-Time Radiance Field Rendering,” these technologies enable photorealistic rendering in real time. Recent advances in Dynamic Gaussian Splatting, particularly the research of Kim et al. [2] on “4D Gaussian Splatting in the Wild,” allow for documenting movements and technical gestures with unprecedented precision.
Beyond the Image: Understanding Technical Gestures
These technological advances must be considered in light of fundamental research on the learning of technical skills. According to Blandine Bril [3], craftsmanship cannot be reduced to the simple reproduction of an observed movement. Technical expertise is fundamentally based on “functional preoccupation” – the ability to direct one’s action toward a specific objective while adapting to variations in materials.
As Bril and Goasdoué [4] explain in their work on perceptual-motor learning, acquiring know-how involves developing an “educated attention” to the affordances of the situation – a sensitivity to the properties of the material and the possibilities for action it offers.
Digitization of a glassblowing workshop and several glass objects at Cerfav using Radiance Field methods and 3D GSC rendering. Photo credit and digitization: David Arnaud – GraciaVR
Towards an Integrated Documentation of Know-how
To overcome the limitations of a purely visual approach, an enriched approach integrating additional dimensions is necessary. The work of Sinha et al. [5] on “SpectralGaussians” offers interesting avenues, allowing for semantic segmentation and analysis of captured scenes.
This integrated approach should include:
- Capturing variations and adaptations of gestures under different conditions
- Annotating the functional intentions that guide the action
- Integrating multiple sensory dimensions (tactile, auditory)
- Documenting the cultural and social context of learning
Digitization using volumetric video of the process of re-creating a stained glass window entitled “Tête d’homme dite Christ de Wissembourg” at the Musée de l’œuvre Notre Dame in Strasbourg. ©David Arnaud – Volumetric video: 4d Views.
A Complement, Not a Substitute?
As Bril suggests in her reflections on educational technologies [6], the effectiveness of these tools will depend on their ability to support active, situated learning. These technologies should be used not as substitutes for traditional learning, but as complements that enrich the learning experience by making visible aspects of know-how that would otherwise remain implicit.
Footnotes
[1] Bernhard Kerbl, Georgios Kopanas, Thomas Leimkühler, George Drettakis — “3D Gaussian Splatting for Real-Time Radiance Field Rendering” (2023)
https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/
[2] Mijeong Kim, Jongwoo Lim, Bohyung Han — “4D Gaussian Splatting in the Wild with Uncertainty-Aware Regularization” (2024)
https://proceedings.neurips.cc/paper_files/paper/2024/file/e95da8078ec8389533c802e368da5298-Paper-Conference.pdf
[3] Bril, B. (2002). L’apprentissage de gestes techniques : ordre de contraintes et variations culturelles. In B. Bril & V. Roux (Eds.), Le geste technique. Réflexions méthodologiques et anthropologiques. Éditions Érès.
https://www.researchgate.net/publication/254861653_Le_geste_technique_-_Reflexions_methodologiques_et_anthropologiques
https://www.editions-eres.com/ouvrage/1217/le-geste-technique
[4] Bril, B., & Goasdoué, R. (2009). Du mouvement sans sens ou du sens sans mouvement : rôle des finalités et des contextes dans l’étude de comportements moteurs. Intellectica, 51(1), 273-293.
https://www.persee.fr/doc/intel_0769-4113_2009_num_51_1_1742
[5] Saptarshi Neil Sinha, Holger Graf, Michael Weinmann — “SpectralGaussians: Semantic, spectral 3D Gaussian splatting for multi-spectral scene representation, visualization and analysis” (2024)
https://www.researchgate.net/publication/383090997_SpectralGaussians_Semantic_spectral_3D_Gaussian_splatting_for_multi-spectral_scene_representation_visualization_and_analysis
[6] Bril, B. (2019). Une approche écologique de l’apprentissage des gestes techniques. In M. Récopé & N. Fache (Eds.), La sensibilité à… : analyse de l’expérience et du sens en activité. Éditions des archives contemporaines.
https://hal.science/hal-00984013/document
