The next great face-off

The “TechKnow” team had nicknamed me the “snow queen” after our story on tornado innovations was derailed by terrible weather. So, of course, though we landed in Boston to glorious sunshine, a heavy storm was predicted for the following day. But it certainly didn't cool my enthusiasm for learning about the world of facial recognition technology.  

Boston seemed like a cross between New York and Dublin — a walkable city, consisting of high-rise buildings and warehouses, and beautiful residential terrace rows of houses, with MIT and Harvard around the corner. We drove toward the shoot, located in the snow-covered peaks of New Hampshire. I could tell we were nearing our destination when my ears started to pop.

Animetrics was located high in the White Mountains, in a cute little ski-resort town — an interesting location for a company at the pinnacle of facial recognition advancement. I was expecting to meet a crowd of code writers, but as one key developer explained, a baseline algorithm for facial recognition has actually sprouted many different and varied applications — one of which I was able to experience firsthand upon my arrival.

The office had a camera hanging from the ceiling just inside the workspace, ready to capture the face of anyone passing the front door. My image had already been pre-entered into the company’s database — from pictures found on the Internet — and so the instant the camera clocked my face as I approached, it connected to the software and the door unlatched. No key, no ID, no hassle. I was in!

Having been based in a software house during the four years of my doctoral research, I’ve seen how basic algorithms evolve. Codes are initially written to perform a specific function. The codes are then developed further, so that the function may be applied in different contexts and applications.

This was certainly the case in my own area of expertise. The initial code of Computational Fluid Dynamics software was built based on Bernoulli’s equation. In time, it was modified to allow engineers to visualize not just how air flows over aircraft wings and Formula 1 cars but also how blood flows through heart stents and oil flows through pipes offshore. Same principle, very different applications.

That explains how Animetrics has developed a number of associated applications, the most impressive of which builds 3-D images of a person’s face from 2-D images, which can be especially useful for identifying criminal suspects from partial or low-resolution surveillance footage.

Facial recognition technology involves converting nodal points into binary sequences. Nodal points — concentrated in areas that are highly characteristic, such as eyes, nose, brows, cheekbones and lips — are placed all over the image of a 2-D face. A computer turns those points into a series of binary numbers that are compared with hundreds or even thousands of faces in a database.

This technique is no different from the principles used in other biometric markers, such as fingerprinting and iris scanning, but the strength of the software very much hinges on the capability of other technology — especially the performance of the cameras and databases. My face, after all, was not being analyzed subjectively, it was being read objectively according to a series of coordinates and binary numbers, used to map it in two dimensions.

Installing the technology at entertainment venues seemed like much more fun. We went to Bar FX in Boston, where a facial recognition system was used at the bar’s entry. I thought it was genius. Building a database of faces for people who need to be legally checked anyway, managing your clients, offering smooth admittance and perks for regular customers seemed to be a win-win alternative to the traditionally slow-queuing systems, where human bouncers have to check through a multitude of often fake IDs. And it instilled a comforting additional layer of security in places where violence fueled by alcohol was more likely to occur.

I didn’t want to leave Boston without visiting the adulated campuses of MIT, a university I have always admired and found inspiring. A friend of mine studying there insisted that she show me two things — the Infinity Corridor and the view from the roof terrace of the Media Lab. Though I wasn't there on one of those few days per year when the sun sets in perfect alignment with the corridor, shining golden light along its entire length, it was still magical to be there in person.

En route to the roof terrace of the Media Lab, my friend and I walked into a hive of activity. Hundreds of MIT students were in the middle of a medical hackathon — a weekend-long challenge that connected health care providers, investors and innovators to find technology solutions to global health problems.

Among the ideas put forward, there was even a project that involved using facial recognition technology to diagnose genetic diseases. Genetic diseases often manifest themselves in highly specific facial characteristics — Down syndrome is probably the most well-known example. Here was an entirely different use from the story I had been working on.

My experience at MIT, together with the serious aspects of law enforcement and the lighter application of the biometric bouncer, all came together to prove that FRT has some far-reaching applications. I think this kind of technology will become a prominent feature of our everyday lives. One day soon, we could be pointing our smartphone cameras at strangers to call up all kinds of personal details about them.

The technology could be as scrupulous or as lax as necessary. It was just a matter of changing the settings. But a computer able to recognize a face in the crowd, for whatever reason, certainly has to be acknowledged as a groundbreaking innovation.

To learn more about facial recognition technology, watch "TechKnow," Saturday 7ET/4PT.