Microsoft’s How-old.net: Science behind guessing your age

News
howoldpicture_110459

(MEDIA GENERAL) – The creators of an age-guessing app that’s moving through Facebook feeds like wildfire are sharing their secrets.

How-old.net claims to guess a user’s age based on a photo of themselves they upload.  In a blog posted on how-old.net, creators Corom Thompson and Santosh Balasubramanian said the app was made in just a day using new face detection rules for building software.  The Microsoft engineering team then sent the test app to several hundred company employees.  Within hours 35,000 people from around the world were using how-old.net.

To understand how how-old.net works, you need to know some basic computer tech language.  An API, or application programming interface, is basically a set of rules for building software applications. The APIs how-old.net uses are available online through Bing and Microsoft’s Project Oxford.  The app’s APIs use a set of algorithms to process photos of faces.  Face detection APIs take in the exact location of an image’s eyes, mouth, eyebrows, nose and lips, as well as face dimensions. The app then uses these characteristics to group people, as well as identify one person in two or more photos.

The makers of How-old.net say it is reasonably good at locating faces and identifying gender, but isn’t particularly accurate with age, leading to some entertaining results. The app uses Bing Search to allow users to share their results on social media, so it’s easy to see why it’s become so popular.

Thompson and Balasubramanian say the how-old.net also grabs information about where each picture was uploaded and the type of device being used, but they stress that no photo information is saved.

Project Oxford: http://www.projectoxford.ai/

How-Old.net: http://how-old.net/

App creators’ blog: http://blog.how-old.net/

Copyright 2019 Nexstar Broadcasting, Inc. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.

Don't Miss