Vincent van Gogh: “Self-portrait”
Setting. In 2019, Art Recognition was approached by a prominent van Gogh expert and scholar to conduct an AI analysis on a selection of van Gogh paintings. Among the works submitted was the controversial Self-Portrait housed at the Norwegian National Museum in Oslo. At the time, we were completely unaware of the ongoing research into the painting’s authenticity, which would later be made public by the van Gogh Museum in Amsterdam at the start of 2020, stirring a global sensation.
Learning van Gogh. To undertake this task, we trained our AI System to recognize van Gogh’s distinctive style. Using hundreds of original images from the de la Faille Catalogue Raisonné, we provided the AI with a comprehensive dataset of van Gogh’s authentic works. To further refine the system’s ability to detect forgeries, we also included images of negative examples, such as the infamous Wacker forgeries, but also imitations and works by admirers and contemporary artists. This meticulous training process resulted in one of our most robust and reliable models, bolstered by the abundance of high-quality images of van Gogh’s works.
One of our most precise assignments. When analyzing the Oslo Self-Portrait, our AI classified it as authentic with a probability of 97% — one of the most definitive and precise results our algorithm has ever achieved. This outcome underscored the reliability of our AI model and its capability to tackle even the most debated artworks.
Epilogue. It is important to highlight that our analysis was conducted just weeks before the van Gogh Museum’s official announcement regarding the painting’s authenticity. Reflecting on the situation, it is somewhat regrettable that we did not publish our results earlier, as doing so would have further underscored the reliability of our AI authentication system. Nevertheless, the consistency of our analysis with the museum’s announcement serves as a powerful affirmation of our technology’s accuracy and potential in the field of art authentication.
