Facial recognition technology is getting smarter and faster. Well, kind of. A company that specializes in face recognition software, fittingly titled Face.com, has devised an API that can not only read faces, but also translate the subject's mood, and even guess their age and gender.
Released in an experimental demo form, Face.com's tool is available for a trial run on its site but it's not quite final yet. When ready, developers will be able to use Face.com's API in their own applications. Imagine, for example, a shopping site that offers different products to you based on your age, gender, and mood.
Not only is the tool's presentation still in the works, the results aren't very accurate either. When it comes to reading emotion, the service can't stop spotting smiles. Case in point: even the sad, weepy face of Dawson's Creek actor James Van Der Beek is read as "happy" with 30% confidence. The service also believes Van Der Beek is smiling, with a whooping confidence of 70%.
The API was also off the mark in the age department, pegging Van Der Beek for a male (80% confidence on that one) somewhere around 44 years old but definitely between the wide age range of 21 and 64 years old.
Here in our office, we gave the API a spin (you can see the results below) and it was never completely accurate. In all but one instance, the subject's gender was determined accurately, which doesn't sound damning, but, really, who wants an app telling them they look like a member of the opposite sex? Plus, age estimates were all over the place, adding a few years here and taking off a decade there without much reason. As far as we can tell, the app is fairly confident that any male with a beard must be over 40.
Check out all the results below, then give the API a try for yourself at Face.com, if you're up for it.
Age in picture: 26
Age in picture: 24
Age in picture: 31
Age in picture: 35
Age in picture: 26
Stay in the know with Laptop Mag
Get our in-depth reviews, helpful tips, great deals, and the biggest news stories delivered to your inbox.