An AI Lens on Historic Cairo

Reports show that numerous heritage sites are in danger due to conflicts and heritage mismanagement in many parts of the world. Experts have resorted to digital tools to attempt to conserve and preserve endangered and damaged sites. To that end, in this applied research, we aim to develop a deep learning framework applied to the decaying tangible heritage of Historic Cairo, known as “The City of a Thousand Minarets.” The proposed framework targets Cairo’s historic minaret styles as a test case study for the broader applications of deep learning in digital heritage. It comprises recognition and segmentation tasks, which use a deep learning semantic segmentation model trained on two data sets representing the two most dominant minaret styles in the city, Mamluk (1250–1517 CE) and Ottoman (1517–1952 CE). The proposed framework aims to classify these two types using images. It can help create a multidimensional model from just a photograph of a historic building, which can quickly catalog and document a historic building or element. The study also sheds light on the obstacles preventing the exploration and implementation of deep learning techniques in digital heritage. The research presented in this paper is a work-in-progress of a larger applied research concerned with implementing deep learning techniques in the digital heritage domain.