Cara Face Recognition Transforms Standard Webcams Into Intelligent Sensors

67 1 Loading

SH 115_#3 BIG

The human face is a treasure trove of information. A millisecond after meeting someone, we’ve guessed their general age bracket, gender, mood, and more.

With tech startup IMRSV’s new face detection software, Cara, your home PC and webcam will learn to recognize some of the same subtleties. Using this information, IMRSV hopes to make analog business and advertising as data-driven and personalized as online business and advertising.

Founder and CEO, Jason Sosa, told Singularity Hub, “Website stats are powerful—gender, age breakdowns, age categories, how many impressions you have based on traffic. Cara gives you the same thing, only it’s for a real world space.” Don Draper won’t need two fingers of Scotch to do his magic anymore.

But Sosa hopes business and advertising are only the tip of the iceberg. “Everybody has a different problem they’re trying to solve. We had a million different people contact us for a million different things. So, we decided to make a platform that enables people to leverage this capability.”

Whereas many computer vision programs tend to be expensive, complicated, and power hungry, Cara is simple enough to run on your PC at home. “Our goal was to build it on really lightweight computers. It runs on my Galaxy S3.”

Currently, the version of Cara publicly available works on Windows and Ubuntu Linux, but in coming months, the firm will release it for other operating systems, including mobile OS’s like Android and iOS. For $39.95/month (per camera), IMSRV offers the software and cloud analytics tools to crunch your data.

Cara detects multiple faces up to 25 feet away and notes whether they are a male or female child (0-13), young adult (14-35), adult (35-65), or senior (65+). It also records how much and what kind of attention they’re giving the camera. This includes total duration (time in front of the camera), glances (looking away and back), attention time (facing the camera), and opportunity to see (traffic near the camera).

How does it work? Sosa explains, “We have a mathematical representation of what, for example, a young adult male’s face looks like. And then on a frame-by-frame basis we’re doing a pattern match to look at your face and compare it to this mathematical representation.”

The team’s been busy putting the software through its paces to ensure it is robust. “We try to get a really general cross-section of society. We’ve tested with beards, scarves, hats, sunglasses, bald people, multiple ethnic groups.”

SH 115_#2These days, when Cara looks at a face, it correctly identifies gender 92% of the time and age 80% of the time.

Advertisers may use Cara to create demographically-driven personalized ads on the street  (i.e., based on age and gender). Businesses may implement Cara to obtain and analyze real time business data. For example, this might include how much attention a product display is getting and how many people walk into a store—comparing locations side by side and breaking them down by demographic.

Claire Fahie, Head of US Retail for Reebok, says, “The essential data that we seek, as a retailer and a consumer product company, is found in the actual human response to product displays, including video content in stores, as well as product end­caps, and we now have a powerful tool for gathering this information in real­time.”

Sosa notes the firm also released an API for developers to invent custom applications they’ve yet to fathom. “We can’t wait to see what people do with it.”

SH 115_#1Some may want to use it to make intelligent products—a toy, for example, that can change how it interacts based on who’s in front of the camera and what they’re doing. Still others want to tie it in to Google Glass.

Sosa says, “Privacy is an important discussion to have. We understand that a Minority Report future is coming, and unless somebody does something about it and creates a better experience, there will be no alternative. That’s the reason we exist.”

Cara was designed to record enough of the critical raw data to fulfill the service’s prime goals—without recording images or personal data. In other words, clients don’t have to worry about IMRSV’s commitment to safeguarding personal information because the firm never has it to begin with.

The terminology differentiates face detection tech like Cara from face recognition technology, which seeks to record, compare, and analyze traits to identify individuals—useful in security applications and less comfortable for many people.

SH 115_#4Sosa is careful to distance his tech from such invasive methods, “We’re describing Cara as a perceptive computing platform. Perceptive computing is a way to differentiate it from any kind of security application, because that’s not what it is.”

Sosa compares real world advertising to the early days of the Internet, when it was a riot of pop-ups and flashing banners vying for consumers’ attention. Cara and other intelligent real world sensors like it could change all that. Cleaning up brick-and-mortar advertising and aiming it more accurately and efficiently.

And after that? “Right now, our computers are pretty basic. But soon computers will look back at you and understand your emotional expression; they’ll know how many people are in the room. Our everyday objects and spaces will understand people better. We all wear our emotions on our face, so what better way to enable that?”

Image Credit: Maik Pereira/Flickr (feature, banner), Keoni Cabral/Flickr (body), Terinea IT Support/Flickr (body), xlibber/Flickr (body)

Discussion — One Response

  • Robert Schreib May 21, 2013 on 6:16 pm

    Heck, ‘The Machine’ on that CBS TV show “Person Of Interest” does all that already! What if they got IBM’s ‘Watson’ to help out in this thing?