Our article "A Neural Network Looks at Leonardo's(?) Salvator Mundi" has been published in the December 2021 issue of Leonardo. Be sure to check out the accompanying tutorial animation!
We are pleased to announce the issuance of our first patent, US 11,195,060, which covers our techniques for producing probability maps. Our classification algorithm, Salient Slices, is and always will be open source.
Noah Kravitz interviewed Steve for NVIDIA's December 1, 2021 AI podcast!
Andrew Ng, the co-founder of Coursera, featured our work in the November 3, 2021 edition of The Batch.
Steve’s article “State of the Art” has been published in the September 2021 issue of IEEE Spectrum (pp. 26-31). "Portrait of the Portrait Sleuths," with a bit about Andrea and Steve's backgrounds and the inception of our project, appears on page 2 of the same issue.
In an August 27 blog post, Brian Caulfield of NVIDIA describes how one of that company's GPUs helps us in our art authentication projects.
Steve's essay The Work of Art in an Age of Mechanical Generation is now available on Leonardo's website. The essay will appear in a forthcoming print edition of the MIT Press journal.
On March 31, 2021, Boston College hosted our virtual presentation "Complementing Connoisseurship: Artificial Intelligence Looks at Painting." We were guests of the Boston College Art, Art History, and Film Department, which had invited us to discuss our work using artificial intelligence to help authenticate paintings. The McMullen Museum of Art hosted the event, during which we discussed how we're using convolutional neural networks to extend scientific analysis to a painting's visual features. In our talk, we explained how we developed our system, using Rembrandt portraits as a test case, and then applied it to works by other artists. We considered the interesting question of how involved Leonardo da Vinci actually was in the execution of "The World's Most Expensive Painting" -- the Salvator Mundi. Finally, we expressed hope that our system might be a valuable addition to the "toolbox" -- which includes other scientific techniques, connoisseurship, and provenance research -- that experts use to explore the lived history of a painting.
In her January 26, 2021 article Disarming new findings on Leonardo's Salvator Mundi, The Art Newspaper's editor Alison Cole considers our system's analysis of the painting alongside the findings of Louvre experts who examined the panel. The mutually supporting findings suggest that Leonardo's initial conception of the work might have been very different from the painting we see (or would see, if we could) today.
Our article "A Neural Network Looks at Leonardo's(?) Salvator Mundi" has been accepted for publication in Leonardo (MIT Press). Pending publication of the article, it is currently available on arXiv at https://arxiv.org/abs/2005.10600 and on the journal's Just Accepted page.
Our article "Salient Slices: Improved Neural Network Training and Performance with Image Entropy" has been published in the June 2020 issue Neural Computation (MIT Press). The article is also available on arXiv.org at https://arxiv.org/abs/1907.12436
On November 10th, 2019, Steve presented The Robots Want to Look at your Rembrandt as a panelist in The Convergence of Artificial Intelligence on Art at the 2019 Art Appraisers of America National Conference in New York City.