If you run a website, you have access to a potential wealth of data about your visitors: their location, gender, approximate age, and maybe even their names, with the help of a simple cookie. If you run a retail store or a mall, that same data must be collected far less efficiently by an army of kids with clipboards. But what if stores could count patrons automatically, tallying up basic demographic data through a webcam installed at the register?
That’s what New York-based IMRSV is trying to do with Cara, cheap face detection software that can scan faces up to 25 feet away and determine age and gender with around 90 percent accuracy using a standard webcam. The technology can be used in stores, advertising, and art exhibits.
Cara is already installed on the wall of shoes at the back of the Reebok store on Fifth Avenue in New York City, where the store is using it to collect data on which customers are lingering at the shoe wall and which ones are leaving. "We're always trying to understand and learn as much as we can about our consumers so that we can create better products for them and better experiences for them," said Inga Stenta, a marketing executive at Reebok.
"We honestly see this as something that’s going to be on every register."
If the experiment goes well, Reebok could install a responsive advertising display that would react differently to different customers. If a man picked up a shoe, the display might play a video about the technology behind it, or an advertisement that featured male actors, Stenta explained. This could potentially be installed in Reebok’s stores around the world, she said.
"We’re tapping into 90 percent of where all commerce happens, and that’s the physical world," IMRSV founder Jason Sosa told The Verge. "We honestly see this as something that’s going to be on every register, in every vending machine, on possibly every billboard that’s out there."
It’s taken the startup more than two years to release its custom-built software. IMRSV conducted extensive testing across six countries to ensure that Cara could pick up people in beards, scarves, and hats. The software is accurate up to 93 percent of the time, the company says. IMRSV envisions a wide range of possibilities for Cara, including watching audiences during live performances and monitoring whether drivers are looking at the road.
Cara does not currently have the ability to detect facial expressions, but eventually the company hopes to add that functionality. One day Cara could power a toy that, for example, could smile back at a child.
Imagine a toy that can smile back at a child
Privacy is an obvious issue when you’re building a technology that obsessively tracks people in the service of capitalism. Cara doesn’t violate people’s privacy because it only tracks people in aggregate, IMRSV maintains, and it does not store images, only data. So while a Cara-powered billboard might switch from skirts to Chinos as a man walks by, it will not be addressing him by name and asking how those assorted tanks worked out, as in the movie Minority Report.
"We're looking for faces or the pattern of what a face looks like," Sosa said. "We’re not identifying that face, we’re just saying, ‘This is a face,’ and we’re able to identify it as a male face or a female face. But we're not doing any kind of personal identity, that's the next step, which is facial recognition. It's very different from that type of technology."
Off-the-shelf face detection software is becoming more common, but Cara is one of the most affordable at $39.95 per camera. It’s easy to download from IMRSV’s website and install, even on $150 computers and cheap webcams. IMRSV is also providing a free API so developers can build apps using Cara.
Face detection software is becoming more common. Intel’s Audience Impression Metric Suite is similar to Cara, but it’s more expensive and it’s designed specifically to be embedded in signs and ads. SceneTap uses face detection slightly differently. The app lets users check out how many people are in a bar — and of course, the male-to-female ratio.
As offline data-gathering gets more sophisticated, however, face detection may be replaced by more powerful technologies based on facial recognition and iris recognition.
Face detection may be replaced by more powerful technologies
"Certain goals or tasks could be satisfied merely with face detection and do not require face recognition," Dr. Alesandro Acquisti, a privacy expert and associate professor at Carnegie Mellon, said in an email. Dr. Acquisti conducted a study on the potential for using social media profiles and facial recognition software to identify people.
"Once the data necessary for face detection has been gathered, and stored somewhere, face recognition is — technologically speaking — just one little step away, not a giant leap away," Dr. Acquisti said. "Hence, it seems likely that services and data initially aimed at mere face detection will slowly (or not so slowly) creep into full-fledged face recognition."
IMRSV says it is not interested in upgrading to facial recognition, only in fulfilling the low-end market for face detection. If it sticks to that promise, it may keep clear of privacy controversies at the expense of its own success.