On 26 March 2026, Laura Willot will defend her thesis entitled:
Structuring, enriching, and exploring photographic collections for the spatio-temporal monitoring of heritage restoration sites
📅Thursday 26 March 2026 – 2.00 pm
📍 Salle G. Morin – building N’ – Campus CNRS Joseph Aiguier – 31 Chemin Joseph Aiguier – 13009 Marseille
How can I attend? By filling in the form available via this link
https://framaforms.org/soutenance-phd-defence-laura-willot-1770667255
Thesis committee:
Supervisor : Dan Vodislav (ETIS, CYU)
Supervisor : Valerie Gouet-Brunet (LaSTIG, IGN)
Supervisor : Livio De Luca (MAP, CNRS)
Tutor : Adeline Manuel (MAP, CNRS)
Reviewer : Gilles Gesquiere (LIRIS, ULL2)
Reviewer : El Mustapha Mouaddib (MIS, UPJV)
Committee member : Florence Clavaud (Archives Nationales)
Committee member : Dimitris Kotzinos (ETIS, CYU)
Committee member : Angela QUATTROCCHI (UNIRC, Italia)
Résumé de la thèse :
This thesis explores the use of digital tools to make the most of image collections acquired during heritage restoration projects, using the documentation of the restoration of Notre-Dame Cathedral in Paris following the fire in April 2019 as a case study. During this project, a vast number of photographs were taken and subsequently enriched with various contextual metadata (date taken, camera position within the cathedral, description of the visual content, etc.), thus constituting a valuable resource for documenting the restoration process. However, their sheer volume makes any manual exploration impossible.
The approach we have developed is based on the concept of distance between images, calculated from their metadata (dates, spatial coordinates, etc.), thereby facilitating the exploration of the corpus. This approach is divided into four stages: Understand, Structure, Explore, Enrich. The first stage (Understand) involves analysing the data and its representations, by studying the tools and context of their production, to define a graph data model that integrates the various connections between metadata and images. The second stage (Structure) focuses on data management, using a graph-oriented database, and the development of deep learning models to generate new data representations. The third stage (Explore) utilises data visualisation (word clouds, dimensionality reduction) and the concept of distance (nearest neighbour search, clustering) to explore the corpus. Finally, the fourth stage (Enrich) is built around a general framework, unifying the previous tools and stages to produce new knowledge.
This work paves the way for the exploration and partial enrichment of large heritage image corpora. Further work will be required to deepen the production of knowledge from these collections.