fbpx
News

An artificial intelligence has determined how to up your selfie game

How to take a better selfie

If you’ve long questioned why your selfies get fewer likes and shares than the ones your friends post (and let’s be honest, who hasn’t?), then wonder no more: a computer scientist and the artificial intelligence algorithm he’s developed, have formulated several tips for taking better self-portraits.

In order, those tips are:

1) The face should occupy one-third of the image.

2) The image should be cropped at the forehead.

3) Extra saturation in and around the face does not hurt.

These tips may seem unintuitive, but they’re actually excellent advice: ask a professional photographer for beginner’s advice on how to capture a stunning image and they will inevitably mention the rule of thirds and how important it is to fill the frame. The bot was also partial to selfies that took advantage of filters and borders. The latter likewise makes sense, as using a frame within a frame is another compositional trick.

It also helps a lot if you’re woman, especially if you tilt your head. When it came to photos with men, overall the system didn’t think male selfies were particularly compelling; in fact, a photo that includes a male is not featured in the AI’s list of top 100 selfies. That said, male selfies that clearly show the head and shoulders did better than their peers.

To develop the system, lead developer Andrej Karpathy, a researcher who worked for DeepMind, the artificial intelligence startup Google purchased in 2014 for more than $500 million, fed a neural artificial intelligence network 20 million photos tagged with the selfie hashtag. Initially, the AI was told to look at each photo’s social stats — likes and shares. It then developed its own system that called upon more than 140 million parameters, including things like shape, symmetry and lighting, to sharpen its algorithm.

A Twitter bot set up by Karpathy will run one of your selfies through the system. You can also find out more about the AI on his blog.

[source]GitHub[/source]

MobileSyrup may earn a commission from purchases made via our links, which helps fund the journalism we provide free on our website. These links do not influence our editorial content. Support us here.

Related Articles

Comments