There is an old story about the French artist Édouard Manet. He advised younger artists that, when painting a bunch of grapes, they should not paint each one individually.
Instead, they should focus on how their eyes perceived the bunch and replicate this natural process in their art. Did their eyes automatically hone in on each singular grape, or did they take in a general impression of the collective?
As the painter famously said, “There are no lines in nature, only areas of colour, one against another.”
Almond Cake with Grapes and Peaches, Édouard Manet
How does this affect visual search and art?
Visual search technology is revolutionising a number of industries today. It can identify specific manufacturing parts or items of clothing in a matter of milliseconds. As a result, consumers can turn their smartphone cameras into visual exploration tools.
In theory, visual search has the greatest growth potential in areas where language fails to hit the target. Art falls squarely within this category.
It would be more effective to search for art works based on moods, themes, or styles, than clunky keywords. Technology that can combine these semantic clues with in-depth visual analysis of paintings can unite customers with artworks they would never otherwise have discovered.
But as Manet said, there are no lines in nature, and object detection needs defined perimeters.
Some attempts to bring visual search have successfully tackled this initial challenge. The Magnus app delivers “Shazam for paintings” by identifying works and artists through a smartphone image.
Magnus app
The next frontier
The above is clearly a useful application of visual search, in specific contexts. But how can the same technology apply to ecommerce?
There is a clear customer problem to solve in the $60 billion global art industry. As people’s buying behaviours move online, it becomes more difficult to browse for inspiration. The traditional ecommerce website templates flatten art into just another commodity.
What if the customer could show a picture of a room they wish to decorate? Artificial intelligence could then suggest works that would complement their existing furnishings.
What if the customer had a picture of an image or pattern they like? Visual search could then match this to relevant items in the company’s inventory.
This should be interactive, to take advantage of what digital technology can enable. Saatchi Art tried out this approach when it added a new “Browse” function. Users click on items they like and this immediately influences their recommendations.
How Cadeera brings visual search to the art world
In truth, these developments still fall short of the human, curatorial touch.
At Cadeera, we are developing an ontology for the furniture and art industries that will bring this sophistication to the digital art world. We combine the latest computer vision and natural language processing techniques with domain expertise. This adds scale to personalised recommendations and bridges the gap between consumer intent and ecommerce search results. Cadeera offers services in data quality management, recommender systems, and multimodal search.
