Leveraging AI for the Analysis of Historical Monuments and the Processing of Cultural Heritage Data

This thesis presents the research and development of Machine Learning and AI techniques for the analysis and processing of cultural heritage data. The main objective is to design a point cloud segmentation application for cultural monuments, addressing the challenges posed by large and heterogeneous data. Two approaches are explored: a machine learning classifier trained on a partially segmented point cloud to label the remaining points, and neural network models trained on annotated datasets to segment points into predefined classes. Despite limitations caused by the scarcity of labeled data, both methods proved effective and are integrated into a user-friendly interface, enabling users to load point clouds, select an approach, adjust parameters, and execute the segmentation. Additionally, a state-of-the-art survey on text recognition in handwritten manuscripts is conducted to assess current AI/ML tools for extracting and organizing text based on relevant keywords. Although this work remained in the research phase, it provided valuable insights into existing technologies and future directions. In general, this work combines applied machine learning development, exploration of emerging technologies, and integration of practical tools, contributing to the digitization and processing of cultural heritage assets.