A Machine Learning-based Model to Predict the Cap Geometry of Anatolian Seljuk Kümbets

Abstract

The funerary structures known as kümbets emerged as a unique typology during the Anatolian Seljuk period (1077–1307). The term "kümbet" refers to a monumental tomb that has a tetrahedral, polyhedral, or conical cap. Although the majority of Anatolian Seljuk kümbets underwent renovation work in the 20^th century, a lack of guidance and insufficient documentation has resulted in very few of them retaining their original characteristics. To support the decision-making processes of experts in future renovation work, this study introduces a machine learning (ML)-based model that predicts the cap geometry of kümbets through the use of section drawings. The model development process begins with the determination of the methods to be employed (Pix2Pix and SSIM). This is followed by data collection, data preparation and refinement, and the training of the machine learning model. Finally, there is testing and validation of the model. The results of both a two-step validation process and objective evaluations show that the ML-based model presented in this study has the potential to use section data to provide predictions of the cap geometries of kümbets.