Have you ever wondered why the photos that you share on Facebook or Instagram don’t get as many likes as other people? Well, thanks to the work of MIT’s illustrious Computer Science and Artificial Intelligence Lab (CSAIL), there is now a computer algorithm can now tell you how popular your photos will be — and more importantly, how to make them more popular. You will be unsurprised to hear that the algorithm discovered that sex definitely sells — posing in a swimsuit or miniskirt virtually guarantees a bunch of likes/shares. Interestingly, the software also found that bold colors such as red, blue, yellow, and pink also garner more likes.
The research, carried out by Aditya Khosla of MIT’s CSAIL with friends from Ebay and DigitalGlobe, uses computer vision and machine learning to answer the age old question “What makes an image popular?” The researchers began with a massive corpus of 2.3 million images from Flickr, and then used deep learning (neural networks) and analysis of social cues to discover what exact combination of image features and social context makes an image likable. [Research paper, PDF]
First the software looks at high-level features, namely objects. Modern computer vision algorithms are very good at picking out and identifying objects in images. Unsurprisingly, objects such as miniskirts, bras, and guns have a strong positive impact on an image’s popularity. Most cuddly animals score a “medium positive impact” rating, while wild boars, dull-looking items of food, and other everyday objects create a “low positive impact.” Spatulas, plungers, laptops, and golfcarts actually create a negative impact.
How patches of color affect the popularity of a photo. Bright colors do well, while muted blues/greens do not.
There’s more to a photo than its semantic meaningful features, though — things like color, gradient, and texture also play a big role in deciding a photo’s fate. It turns out that both color gradients and patches of bright color can significantly boost an image’s popularity. In the graph above, you can clearly see that reds and yellows tend to do a lot better than muted blues and greens.
Finally, the algorithm looks at social cues — how many friends the uploader has, how many total views/likes the uploader has across all of their images, how long the uploader has been using the service, etc. It turns out that having a large number of friends/contacts, tagging, and the mean number of views, are key factors in deciding how popular an image will be.
The most shared photo of all time: Ellen’s Oscar selfie. If only Bradley Cooper was wearing a bikini. And Jennifer Lawrence was holding a composite longbow.
How to make your photos more popular
If you want to find out how your own photos score in terms of popularity, you can actually head along to Khosla’s website and try the algorithm out. It doesn’t perform any of the social analysis, though — just the high- and low-level features, such as the presence of objects, or specific patches of color.
Do not upload boring (or disgusting) looking food if you want your photos to be popular. Martha Stewart uploaded this one: 3 million followers, and yet it was only retweeted 24 times.
Alternatively, I’ve taken Khosla’s findings and turned them Sebastian’s 5-Step Guide To Uploading A Popular Photo:
- Upload photos of people, not landscapes.
- If you do upload photos of scenery or landscapes, make sure they contain pretty gradients and bright colors (sunsets!)
- If you’re trying to take the perfect selfie, you should take most of your clothes off and hold some kind of firearm.
- If you insist on uploading photos of food, at least try to make sure it’s bold, colorful food — not drab, pasty food like guacamole.
- Be sure to tag/hashtag your photos adequately and appropriately (though you might hate them, generally the more tags you use, the more views you’ll get).
