Everybody knows that feeling when you post a photo online and it's universally ignored – an image forever consigned to a lonely, embarrassing and very public graveyard. But, thanks to a team of MIT scientists and researchers, that may never happen again. They've created an algorithm that predicts how well-received your Instagram selfies will be. So put away that photo of you dressed up as Snoop Lion on Halloween and check first – remember, there is a good chance that what you think is a surefire hit, ain't.
Led by MIT doctoral candidate Aditya Khosla, the researchers came up with an algorithm that predicts the online popularity of a picture on a scale from zero to 10. By analysing almost three million images on photo–sharing site Flickr, they came to the conclusion that they could forecast what was going to be hot depending on the content of the image and its context.
The algorithm also takes into account colours, number of objects in the shot, as well as textures and gradients. Then it scores the photos based on objects of "positive impact". So miniskirts (no shit) and revolvers have a strong positive impact. Guacamole and catamarans possess a low positive impact, while spatulas and golfcarts actually have a negative impact.
We put the algorithm to the test by with this picture from our Instagram account of Mariacarla Boscono playing guitar in a hotel. This snap took home 922 likes – but what does the algorithm say? It scored Mariacarla 4.587 – a popularity score of under 50%. Sadly, it looked like we should have put her in a miniskirt – or, as the research advises, we should have put her in a more colourful dress. But it is Mariacarla playing a guitar, so give us a break.
You can try your luck with the site here.
Follow Thomas Gorton on Twitter here @angstromhoot