The algorithm was presented as part of a paper entitled “Neuroaesthetics in Fashion: Modeling the Perception of Fashionability” at a conference held in Boston last month, according to a report from Science Daily. Edgar Simo-Serra and Francesc Moreno-Noguer from IRI teamed with Sanja Fidler and Raquel Urtasun from the University of Toronto to make the project a reality.
The model was built using a dataset of fashion-themed posts found on the Internet. Images of fashionable clothing and the captions they were given were cross-referenced against the social media response to each entry. Individual garments were judged as fashionable or not fashionable based on the amount of likes they received on social media.
Basically, the method boils down to posting a picture of an outfit on Facebook and seeing whether people love it or hate it — which is, in of itself, a decent summation of how fashion works. However, the project can claim some authority on what to wear thanks to the scale of the sample being used, with data from over 144,000 posts being taken into consideration.
The intention is that the system will be able to advise users on fashion choices thanks to the data it has accumulated. What remains to be seen is whether anyone will subject themselves to having a computer give them tips on how to dress in the morning. If the idea of fashion can really be boiled down to a statistical algorithm, wouldn’t we all end up wearing the exact same outfit?
Editors' Recommendations
- Gmail is getting a major visual change, but you can opt out
- Next-gen GPU prices will all come down to crypto. Should you wait to buy?
- How will we know when an AI actually becomes sentient?
- Microsoft quits its creepy, emotion-reading A.I.
- The new ways Meta will pay you to make content for Facebook and Instagram
