Cultural heritage is an integral element of today’s society. It is crucial to the creation of a common identity and the connection to people with similar backgrounds. Historical cultural heritage monuments such as churches and temples are hard to document and organise in digital libraries because of aging, destruction, and geometric complexity. The ability to reason about the structure and style of a monument is therefore of great importance to cultural heritage experts, as it can assist them in analysing and cataloguing it. Different from previous work in the field of architectural history which is confined to 2D studies, this proposal proposes an innovative framework based on artificial neural networks for learning the structure and style of historical monuments from 3D data.

The framework is composed of a neural network that can segment monuments into architectural components, a neural network that can detect the style of a monument and a shape grammar extraction and comparison method. The neural networks will be trained on an annotated 3D monument dataset of historic monuments, and used together with the shape grammar to enable a software package for: organizing monuments according to style; observing the parts of a monument and highlighting its main stylistic influences; analysing and comparing the design rules of monuments. 

The objectives of this proposal are: to develop the neural networks and shape grammar; to collect annotated monument 3D data to train and test them; to develop and evaluate the software package for assisting cultural heritage experts in their documentation workflows. The successful outcome of this project, ensured by the partners’ experience in computer graphics, geometry processing, machine learning and architecture, will have a definite impact on their scientific excellence and expertise, as well as society in general, through its application to the digital cultural heritage field.

This work will be carried out within an interdisciplinary environment involving the University of Cyprus, The Cyprus Institute, the Universitat de Girona and counseling regarding the ANNs from the University of Massachusetts Amherst.