Cultural Heritage Image Classification

Abstract

Image classification in cultural heritage context represents one of the most important tasks in the process of digitalization. In these terms, classification can be particularly challenging due to a high number of different image categories, feature variability, and the need for high reliability. Recent research shows that various machine learning techniques can be utilized for image classification purposes and that algorithms such as artificial neural networks, decision trees, and support vector machines are able to obtain high performances. This chapter explores the deep learning architectures used for classification models. Furthermore, we are conducting research on the image classification of Eastern Orthodox cultural heritage, which may assist in the future process of digitalization. In particular, we created a dataset, as such to our knowledge does not exist, containing images of Eastern Orthodox cultural heritage, namely frescoes and sacral objects. The dataset is available for the public, and it represents an additional novelty of this research. Different classification methods are applied to the dataset with the aim of finding the most suitable configuration that will yield high classification performance.