Group of worldwide researchers create AI-powered software to detect fake Raphael masterpieces

A collaboration of worldwide researchers interested in art has created innovative software to detect fake art masterpieces. Their software utilizes artificial intelligence and machine learning techniques to distinguish art painted by different artists. They have used this software on Raphael’s Madonna of the Rose and found that one section might not have been painted by Raphael.

Art forgery is a craft in itself, and some have mastered the ability to create pieces that look like the work of masters. However, minute differences in how one holds a brush and applies pressure to paint strokes exist. These telltale signs are what the software picks up on to detect forgeries.

Technically, the AI/ML software first looks at a set of authentic art pieces created by various artists using a ResNet50 model to identify distinguishing features. Next, the software uses SVM training to correctly identify which art piece belongs to Raphael using the ResNet50 features. Finally, fine brush stroke details are analyzed using 4 different edge detection operators to learn the distinguishing minutiae of a Raphael brush stroke.

The researchers tested their software on Raphael paintings. While Raphael was correctly identified as the artist on most, the Madonna of the Rose was identified as possibly being painted by someone else. Closer analysis revealed that only the male face might have been painted by someone else while the remaining by Raphael.

However, their software could not identify the correct artist of a painting approximately 2 to 3% of the time, so this determination might not be true. Until software accuracy is 100% perfect at detecting fakes, read about art forgery in this book at Amazon and try your hand at copying Raphael's Madonna of the Rose.