Algorithm Predicts How Popular Your Photo Will Be Online

(By Anthonia Akitunde)

As social media and content have become more image driven, more attention has been paid to the power of a photo: While some images get millions of views, others are completely ignored. Even from the same users, different photographs receive different number of views. This begs the question: What makes a photograph popular?

The rules of social media make popularity a nebulous concept. The number of friends and fans who “like” you doesn’t matter if no one is actually engaging with your posts via likes, retweets, favorites, etc. But what if there was a way to find out how popular a post would be before you launched it into newsfeeds everywhere?

That day may already be here for Instagram users. MIT grad student Aditya Khosla and his fellow researchers have released a research paper detailing the factors that make a photo popular. As social media and content have become more image driven, the team writes, more attention has been paid to the power of a photo: “While some images get millions of views, others are completely ignored. Even from the same users, different photographs receive different number of views. This begs the question: What makes a photograph popular?”

The paper’s authors used artificial intelligence to analyze 2.3 million photos on Flickr for content (e.g., color, the subject of the photo, texture) and social context (how many followers a user has, how many tags a photo has, the number of photos in the account). Some of the takeaways from their research:

  •  Cooler colors (greens and blues) tend to do worse than warmer colors (more reds). “This might occur because images containing more striking colors tend to catch the eye of the observer leading to a higher number of views,” the paper states. Brighter colors are more fetching than muted, natural tones.
  • Photos with objects are more popular than ones without objects. The researchers were even able to break down the impact different types of objects have. Swimwear (“maillot” and “bikini”), “perfume” and “cup” create a “strong positive impact,” while a “negative impact” is felt when a “laptop,” “spatula” or “plunger” appears.

Khosla and his colleagues have put their findings into a photo popularity algorithm that users can test for themselves. Upload a photo, and the site will compute a popularity score, which, when 2 is raised to that number’s power, is supposed to tell you how many views your image will get. (For example, a river scene got a popularity score of 4.3, which means it will be 16 times a day.) The site advises that it’s not “entirely accurate,” and that people should pay more attention to the popularity score itself—if one image gets a 6 and the other a 4, you should go with the more popular post.

While the findings apply for all photos, they may be of special interest to businesses that use Instagram, which traffics heavily in images. Though owned by Facebook, Instagram doesn’t have the same filters, fees and algorithms that make one type of post more visible than the other; it offers up a near-constant, real-time stream of posts from the friends and brands you’re following. (It was even named as a better option to Facebook in food delivery company Eat24’s now famous breakup letter.)

(Source: Openforum)

“Opinion pieces of this sort published on RISE Networks are those of the original authors and do not in anyway represent the thoughts, beliefs and ideas of RISE Networks.”

RISE NETWORKS

"Nigeria's Leading Private Sector and Donor funded Social Enterprise with deliberate interest in Technology and its relevance to Youth and Education Development across Africa. Our Strategic focus is on vital human capital Development issues and their relationship to economic growth and democratic consolidation." Twitter: @risenetworks || Facebook - RISE GROUP || Google Plus - Rise Networks