When you’re standing cluelessly in front of a mirror in a new outfit facing three contradicting text messages from your “Is This Outfit Cute?” Caucus, sometimes you wish you could have an objective third party just tell you: Is there something missing? Are you just being crazy and does it actually work? Or, are you being the other kind of crazy and it’s not working at all? A team of researchers from the Institute of Robotics and Industrial Informatics (IRI) in Barcelona, Spain, is creating a tool that attempts to answer the unanswerable.
In its paper, “Neuroaesthetics in Fashion: Modeling the Perception of Fashionability,” presented at the 2015 Conference on Computer Vision and Pattern Recognition, the team set out to determine whether “fashionability” (determined as how many people “liked” a person’s look on outfit-sharing site Chictopia) could be broken down into certain patterns. For instance, are monochrome outfits considered more or less stylish if you’re wearing all black? Or, is denim only fashionable when you wear them with heels (and live in a Western European nation, and are under 25, and your jeans could be described as “skinny”)?
What it found: Some of the strongest correlations were between fashionability and the GDP of the country the outfit is found in. The richer the country, the higher the fashionability. And to confirm your worst suspicions, it also found that the younger the wearer was, and the more conventionally attractive, the more fashionable her outfit was considered to be.
If you have concerns that this algorithm takes away the spontaneity, creativity, and anti-formulaic thinking that makes for truly great style, rest assured that the humans who created it get that, too. “We are not attempting to tell people that their individual style is wrong, and that there is some sort of global fashion they have to adhere to,” says Edgar Simo-Serra, one of the researchers on this project. “The idea here is to provide people with advice to help them with their decisions. We know that fashion is something that is not inherently objective. However, in order to be be able to learn article-intelligence models that can make decisions, we need to provide some sort of objective measure.”
Simi-Serra identified a few common suggestions that the algorithm typically made to improve an outfit’s fashionability: “We’ve seen a preference towards heels and black garments. I’m guessing it’s because heels, in general, are associated with fancier, elegant clothing, and black garments are, in general, easier to match in an outfit.”
Unfortunately, the team was unable to fulfill our requests to test out a few images to rate fashionability (we wanted to test some hypotheses of our own about the same clothes on plus- and straight-size wearers, and challenging — but cool! — fashion trends, as well as one particular outfit that I needed some help with), but it's currently working on an app that you’ll be able to play around with yourself. Until then, continue utilizing your Caucus, take risks and be daring, and put on a pair of heels if you’re really stumped.
AdvertisementADVERTISEMENT