Home Search optimization How a startup is using a facial recognition engine during COVID

How a startup is using a facial recognition engine during COVID


At the start of the COVID-19 pandemic, Sentry Interactive, an organization that provides contactless entry to businesses, shifted from just helping people enter buildings and began performing COVID-19 screenings.

With the goal of improving customer interaction with access control and visitor management, the vendor seeks to create “an experience where everyone who enters a building can feel really good,” Kent said. Kubie, vice president of sales and marketing at Sentry.

Formerly known as the Sentry Kiosk, the Austin-based startup’s Sentry interactive tower recognizes people’s faces and is able to not only identify who they are, but also their reason for entering a place.

The Tower is a platform that includes cameras, sensors, microphones, and a facial recognition engine that engages users trying to enter a facility or building with a digital receptionist who talks to people.

At the start of the pandemic, Sentry was to include a facial recognition engine sensitive to face masks.

CyberLink Face Me

The kiosk company turned to CyberLink’s FaceMe. CyberLink is a provider of multimedia software and facial recognition technology.

FaceMe is an AI-based facial recognition engine that detects face masks and supports different operating systems including Windows, Linux, Android, and iOS.

CyberLink developed FaceMe in 2018 as a software development kit (SDK) and enabled deep learning on the platform using Intel’s OpenVino (Open Visual Inference and Neural Network Optimization) toolkit . The toolkit enables developers to more quickly infer deep learning models and build computer vision applications.

FaceMe identifies faces as well as the inappropriate use of masks, even when someone isn’t looking directly at the camera. When a person’s face mask does not cover their nose and mouth, the engine is able to identify it and ask them to correct it.

“We have found no other [facial recognition engine in] the market that was sensitive to this stuff, which we thought would be very beneficial to our clients as they were from a risk and compliance perspective,” Kubie said. “They had to make sure everyone who came into their organization at a facility was following all the right protocols.”

FaceMe is also easy to integrate, he said.

Because FaceMe is available as an SDK, Sentry’s IT pros could embed the engine into the tower without coding, he said. And the product is also easy to run in the cloud and flexible enough for developers to make changes as needed.

“If we need to make a change to the platform like we’re coming up with a new model and a customer wants that new model, we don’t have to recode everything for that model,” Kubie said. “We can run this on multiple platforms. And if someone starts with one and wants to migrate to another, it’s very transparent for them.”

FaceMe Platform

On April 27, CyberLink released a new web version of the FaceMe platform. The platform is aimed at developers looking to integrate the engine into any system, workflow, or IoT framework.

It includes facial recognition APIs based on the Hypertext Transfer Protocol, as well as 1:1 face matching, 1:many face search, video analytics, image quality check and people grouping .

The platform also includes a central management console that allows developers to view CyberLink’s records of face matches and face searches.

Facial recognition, privacy and prejudice

Because it uses a facial recognition engine, Sentry is sometimes asked what it does with the data of faces detected by the engine, according to Kubie.

The company tells customers that the engine doesn’t capture the image of the face, which is another reason FaceMe turned to Sentry, he said.

“There’s absolutely no way to recreate your face with this algorithm because it’s encrypted,” he said. “It’s data points on your face versus your photo of yourself so you can’t be spoofed.”

“We do not store an image of you. We store an encrypted algorithm that if someone got it, they couldn’t do anything with it,” he continued.

And as with most facial recognition and biometric systems, skeptics criticize racial bias. Although Sentry claims to have never faced such criticism from customers, CyberLink said that someone’s skin color does not affect the engine because FaceMe includes a database of images from around the world. entire.

“That has never been an issue for our SDK,” said CyberLink VP of Business Development Terry Schulenberg. “The only thing I’ve ever seen is misidentifying a man or a woman and knowing whether or not they have a beard. Bad lighting can make an African American woman look dark. has a beard when she doesn’t because of the shade.”

CyberLink charges customers on a per-camera basis for FaceMe. A camera means a license that can range from $50 to $250 per camera.

Sentry did not reveal how much it was paying for the facial recognition engine, but said the technology has not resulted in a noticeable price hike for its customers.