Exploring Artistic Styles in Pieces: A Computational Approach to Artwork Fragment Classification

The restoration of ancient frescoes presents significant challenges, particularly when fragmented pieces from multiple works are discovered intermingled. To restore these frescoes, an essential first step involves accurately grouping fragments belonging to the same artwork. This grouping requires a reliable method to classify fragments based on their stylistic and visual characteristics, facilitating focused reassembly efforts for each fresco. Artworks, including frescoes, are invaluable cultural and historical artifacts that embody the aesthetic, social, and spiritual values of their time. However, their preservation is frequently jeopardized by natural disasters, human activity, and material degradation over centuries. Style classification offers a systematic approach to organizing and studying these fragmented works, enabling effective differentiation of pieces that belong to distinct frescoes. This study explores the use of artificial intelligence, particularly machine learning and deep learning methods, to classify both entire fresco images and their fragmented versions. The classification process considers different levels of fragmentation, breaking images into 10, 20, 40, and 80 parts, to analyze performance across various granularities. Alongside visual data, the analysis incorporates color histograms to extract additional stylistic cues. The proposed approach demonstrates robust performance, achieving significant advancements in grouping fragments by style compared to prior methods. By accurately classifying fragments based on style, this research lays the groundwork for more efficient restoration workflows, where fragments from the same fresco can be identified, grouped, and subsequently reassembled. The findings contribute to both the computational study of cultural heritage and the practical challenges of preserving and restoring fragmented artworks.