A research team from the University of Granada has developed ARTDET, an innovative AI software designed to detect deterioration in easel paintings. This tool automates the analysis of damages such as paint layer loss, streamlining the work of art restorers. The project, led by Francisco M. García-Moreno—member of the MYDASS Research Group and the OnTheEdge R&D Project—employs deep learning techniques to identify and classify different types of damage, optimizing restoration processes.
Automating the Detection of Painting Deterioration
ARTDET was developed in collaboration with conservation and restoration experts Luis Rodrigo Rodríguez Simón and José Manuel del Castillo de la Fuente from the Department of Painting and Restoration at the University of Granada, as well with María Visitación Hurtado Torres—also member of the MYDASS Research Group and the OnTheEdge R&D Project. According to García-Moreno, “ARTDET is not intended to replace the expert eye but to assist in the most labor-intensive aspects of damage documentation, providing restorers with a complementary and accessible tool while reducing the time required for manual techniques.”
The AI was trained using high-resolution images annotated by professional restorers. The software achieved an 80.4% accuracy in damage detection and a 99% confidence level in its predictions. José Manuel del Castillo explains: “Our goal is to provide restorers with a fast and accurate reference point during the documentation process, helping them guide their interventions more effectively.”
A Tool to Enhance Art Restoration
The ARTDET system significantly reduces the time and effort needed to create damage maps, which traditionally required extensive manual labor. “ARTDET offers results that experts can verify and adjust easily, combining the best of both worlds: technology and human expertise,” adds Del Castillo. The software and the dataset containing images of paintings with various types of deterioration are freely available in open access repositories. “We aim for this project to be a starting point for future improvements, both in technical aspects and its application in different areas of art conservation,” concludes García-Moreno.
Collaborative and Open Access Project
This project is a multidisciplinary collaboration between departments of art restoration and computer science. The AI software, ARTDET, and its accompanying dataset are available through open access platforms, promoting collaboration in the conservation field.
Github repository: https://github.com/frangam/artdet
Dataset: https://doi.org/10.5281/zenodo.8429815
Reference
García-Moreno, F. M., Cortés Alcaraz, J., del Castillo de la Fuente, J. M., Rodríguez-Simón, L. R., & Hurtado-Torres, M. V. (2024). ARTDET: Machine learning software for automated detection of art deterioration in easel paintings. SoftwareX, 28, 101917.
DOI: https://doi.org/10.1016/j.softx.2024.101917
