Restoring Paintings With Machine Learning
Pompeii fresco fragments (credit: Project RePAIR)
How do you restore an ancient fresco, painted nearly two-thousand years ago, that now lies shattered in thousands of fragments? You design an international collaboration between archeologists; art restorers; robotic engineers; and machine learning software programmers. RePAIR, Reconstructing the past with Artificial Intelligence and Robotics, was established by the European Union in an attempt to do just that. The EU project aims to restore ancient frescos and other artifacts that once decorated buildings in Pompeii before the Roman town was buried when Mount Vesuvius erupted in 79AD.
According to RePAIR, a primary goal of this cutting-edge technology demonstration is to develop an innovative way to eliminate one of the most labor-intensive steps in archeological research, ie. the reconstruction of shattered objects by hand. Most artifacts from excavation sites around the world have not survived intact and are uncovered in broken conditions. The laborous task of matching the pieces into a restored condition is time-consuming and expensive. If robots could be trained to do this, it will allow archeologists and historians more time for exploration and discovery. Pompeii was a perfect laboratory for RePAIR to demonstrate the potential of their vision.
Reconstructing Pompeii frescos and pottery with AI-trained robots (credit: Project RePAIR)
The program is integrating technologies from diverse academic and industrial fields that typically don't interact. By utilizing cutting-edge bio-mimicry influenced robots; computer vision 3D scanners; machine learning algorithms; and art restoration, the EU program hopes to showcase that reconstruction can be accomplished rapidly and ancient artworks can be recovered which otherwise would have remained shattered. RePAIR is creating an intelligent robotic system that will autonomously process, scan, match, and re-assemble fractured objects using flexible "hands" in a fraction of the time that would have been required by a person.
The multi-disclipinary effort has developed completely novel solutions to reach its lofty goals that potentially could have wider applications. RePAIR is pushing the limits of art preservation, historical discovery, and the use of machine learning in practical, real-world situations. It would be worth a visit to Pompeii to watch the progress as the 2000 year old paintings come back to life. WHB