Image Popularity

The past week’s posts have considered “what makes an image memorable” based on an algorithm developed at MIT. It happens that there’s also a MIT algorithm that examines image popularity. The next few posts will take a look at it and ultimately consider what correlation, if any, exists between memorable and popular images.

 Preview of Coming Attractions

Spoiler Alert! A memorable image doesn’t guarantee a popular image – and vice versa. But that shouldn’t come as a surprise. We knew that without MIT’s confirmation.


The MIT image popularity algorithm, in a nutshell:

  • Is based on analyzing 2.3 million Flickr photos
    • To see which got the most views
  • The correlation between the color, composition and subject was examined
    • To determine an image’s likely (internet) popularity In terms of views

 

Memorable does not equal popular (necessarily)

The first image below is 5X more memorable (76% vs 15%)

but is likely to receive less than 1/2 as many views (7.9 vs 18)

Memorability: 76% Expected views (normalized): 7.9

Memorability: 15% Expected views (normalized): 18

Don’t take either of these algorithms too seriously

They may be indicators, but they’re nothing more

More to come…


Enter your email address to follow this blog and receive notifications of new posts by email.