Artfinder is bringing AI and predictive analytics to the art market. The e-commerce platform aims to connect artists and buyers around the world, independent of traditional gallery and institutional gatekeepers, using a new machine learning algorithm. The startup leverages population-based collaborative filtering algorithms, and graph database firm Neo Technology’s software, Neo4j, to enable bespoke buying suggestions and increase art sales.
Today the art world is beginning to embrace the ongoing digital revolution. Influencers in the field, both individual and institutional, have adopted technology to enable easy access, navigation and convenience to the often inaccessible art market.
Online auction house Paddle 8 has merged with a swiss tech company, The Native, allowing for their online auctions to accept Bitcoin. Sotheby’s has announced its decision to purchase Thread Genius, a tech startup specializing in taste image recognition.
Big data, information and research are becoming the driving tools to further the development, growth and sustainability of new and established art businesses. Leveraging the way some of the biggest players in entertainment have completely reshaped the way we navigate culture—IMDb for movies, Spotify for music and Amazon for books—a new service for art has emerged: Artfinder.
Artfinder is a London-based digital art marketplace that connects artists with collectors around the world. By eschewing the brick-and-mortar, curator-led art world, Artfinder is ushering in a new era of collecting and building collections via its online store.
Jonas Almgren, the site’s Swedish Chief Executive, considers Artfinder the ‘OkCupid for art’. “We’re more of a dating site than an e-commerce site, even though we are setting you up with an artwork,” Almgren says. “But we have the same issues as a dating site: How can you know what you like until you see it?”
Data-driven art recommendations
Timing is a key element of Artfinder’s value proposition. The goal is to match a collector with a compatible piece of art before his or her interest wanes. For buyers with a budget over £500, Artfinder provides personal shoppers to filter the site’s database of 400,000 artworks as per individual preferences.
Branded as Artfinder’s “AI curator,” the company’s AI software, Emma, is a Twitter bot which provides data-driven recommendations that are entirely unique to the individual buyer. These recommendations are powered by a combination of population-based collaborative filtering algorithms and Neo Technology’s graph database software, Neo4j.
The software analyzes the kind of artists a user follows on their Artfinder account, as well as their likes and dislikes on the platform’s browse history. Furthermore tweeting an image to Emma would then allow it to compute similar images of artworks on Artfinder, depending on style and composition of the tweeted image.
Diving deep into image data
Emma is powered by visual search technology, which is also implemented on the site as a “more like this” feature, built on Artfinder’s proprietary DVS (deep visual search) technology. “If you see something you like, you can use the ‘more like this’ function to browse by taste or colour in an informed way, rather than doing research,” Almgren says.
DVS uses LIRE in a framework that can derive similarities between artworks that go beyond traditional visual attribution (such as sunset matches sunset, or Impressionist style matches Impressionistic style), looking at the deeper visual structures that evoke a human reaction, and then find other artworks that evoke a similar response.
“Through Emma and her related technology, we believe we have cracked the two biggest problems in art e-commerce. Firstly, that users don’t know how to describe what they want—text search is no use for a visual medium—and secondly, when selling original artwork, what do you do when the one you want has already sold?” Almgren says.
He adds that artworks are added to the Artfinder database at an exponential level, and thus recommendations are key. “Around 1000 new artworks are currently being added to the site every single day, so even if you’re refreshing our ‘new’ page every hour, there’s lots of great stuff you’re going to miss.”
Emma is also powered by an advanced image recognition technology, used by Danish police to identity criminals via CCTV, in order to provide similar-looking images as one browses the Artfinder database.
The art space has understandably been one of the more difficult to bring into the digital arena, as data is often unstructured, visual and disparate. However, new technology like Emma and Artfinder are working to bring data-driven decision-making to the space.