Artificial Intelligence and Restoration: How Algorithms Transform Approaches to Art Preservation

The use of artificial intelligence and digital technologies expands the capabilities of art historians and restorers, offering new approaches to preserving cultural heritage sites and working with them. The article presents an overview of the world experience of using digital solutions to perform complex restoration tasks related to the recovery, identification and reconstruction of works of art and architecture. Particular attention is paid to projects that have become widely known outside the restoration field. Among them are the reconstruction of fragments of Rembrandt’s «Night Watch», the digital reconstruction of fragments of Leonardo da Vinci’s «The Last Supper» and the restoration of the colors of Gustav Klimt’s «Faculty Paintings». In these projects, artificial intelligence algorithms presented the final image of the lost elements, which were very close to the authentic ones. The project «The Next Rembrandt» is mentioned, which demonstrated to the general public the ability of artificial intelligence to deeply analyze the author’s style and its interpretation. The principle of using neural networks in restoration practice to detect early images hidden under later paint layers is revealed using the example of studying the wings of the Ghent Altarpiece by the Van Eyck brothers. The projects of the British startup Oxia Palus are also considered, combining the capabilities of neural networks to analyze the author’s style and detect the original image under layers of paint, using the example of paintings by Pablo Picasso and Vincent Van Gogh. The article emphasizes the importance of an interdisciplinary approach that brings together art historians, archaeologists and engineers, as in the RePAIR project, aimed at recreating artifacts from ancient Pompeii. An example of the successful use of 3D scanning technologies in archaeological research is presented, such as digital reconstructions of architectural monuments in Palmyra and Mosul. Such significant successes in deciphering ancient texts as the study of scrolls from Herculaneum using tomographic scanning and machine learning algorithms are demonstrated. In addition, suggestions are made about the possibilities in the field of cultural heritage preservation that could be provided by further improvement and integration of algorithms into restoration processes.