A New Algorithm Can Help You Take Better Selfies

A lot of thought goes into a selfie. Is the lighting good, am I having a perfect hair day, are there sesame seeds in my teeth, do I look enough like a duck?

Millennials and confused grandparents need not fear for their selfie parameters anymore, because computer scientists at the University of Waterloo have developed an app to educate people on the art of snapping fantastic self-portraits.

The method revolves around an algorithm that directs users towards the best position of their camera, allowing them to capture the best possible angles. A few early social media adopters might refer to this as “Myspace” angles, but for those of you who began their online life with Facebook or still take photos using tablets, it simply refers to your good side.

One of the best aspects of this new app is that it provides direction. Users won’t just see an enhanced version of themselves, but also the process, so they can learn exactly why a photo is better.

Vogel and Qifan Li developed the algorithm by purchasing 3D digital scans of “average” looking people. They then took hundreds of virtual selfies by writing code to control a virtual smartphone camera with fake lighting. This allowed them to explore composition, from lighting direction to face position and size.

They then posted the results online and crowdsourced thousands of people to vote on the virtual selfies they thought were best, then mathematically modelled the voting patterns to guide users towards an ideal selfie.

Following the virtual selfie came the human-testing period. Subjects took photos using a standard camera, then once more with the algorithm powered app. Selfies were found to be 26 per cent better with the University of Waterloo app, though no promises were given that your Tinder matches would increase by 26 per cent as well.

“This is just the beginning of what is possible,” said Vogel. “We can expand the variables to include aspects such as hairstyle, types of smile or even the outfit you wear.”

Vogel and Li recently presented the work in Edinburgh, Scotland at the 2017 ACM Conference on Designing Interactive Systems.