3D Surface Approximation of the Entire Bayeux Tapestry for Improved Pedagogical Access

Marjorie Redon, Matthieu Pizenberg, Yvain Quéau, Abderrahim Elmoataz; Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV) Workshops, 2023, pp. 1593-1602

The Bayeux Tapestry is an exceptional cultural heritage masterpiece by its size and the finesse of its details. Digitizing it raises a challenge, knowing that it is extremely fragile and thus lasers or invasive techniques are out of scope. In this work, we address this 3D-reconstruction challenge by introducing a pipeline to generate a high-resolution panorama of the Tapestry's geometry. It is based on a deep learning architecture that converts the RGB images of a pre-existing 2D panorama into a 2.5D normal map panorama. With a view to facilitating the Tapestry inclusive accessibility, we further show that coupling our 3D-reconstruction pipeline with a segmentation method allows the affordable and rapid creation of 3D-printed bas-reliefs, which can be explored tactilely by visually impaired people.