Semantic Segmentation using Foundation Models for Cultural Heritage: an Experimental Study on Notre-Dame de Paris

Kévin Réby (1) , Anaïs Guillem (1) , Livio De Luca (1)

1 MAP - Modèles et simulations pour l'Architecture et le Patrimoine

Vision foundation models have already had a major impact on several computer vision tasks. This work aims to study their usefulness in the context of cultural heritage. By utilizing the Segment Anything Model (SAM) we could perform segmentation on Notre-Dame de Paris images. Additionally, we have developed a pipeline that combines various foundation models (GroundingDINO and CLIP) to demonstrate their abilities for semantic segmentation of cultural heritage data.

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Machine Learning [stat.ML]

Kévin Réby, Anaïs Guillem, Livio De Luca. Semantic Segmentation using Foundation Models for Cultural Heritage: an Experimental Study on Notre-Dame de Paris. 4th ICCV Workshop on Electronic Cultural Heritage, Oct 2023, Paris, France. ⟨hal-04275454⟩