Vision Al Blog
A Deep Dive with Professor Marios Savvides on “Safety Recognition”
This week I had the opportunity to interview Professor Marios Savvides, a Professor of Electrical and Computer Engineering (ECE) and founder and director of the Biometrics Center at CMU. In October, Oosto announced a new collaboration with Carnegie Mellon University’s (CMU) CyLab Biometric Research Center, a former Gold winner of the Edison Awards for Applied Technology. This collaboration will focus on early-stage research in object, body, and behavior recognition.
ECE Professor Marios Savvides has an impressive pedigree. Professor Savvides was named one of the “2020 Outstanding Contributors to AI” awards from the former U.S. Secretary of the Army. His research has been focused on developing core AI and machine-learning algorithms that were successfully applied for robust face detection, face recognition, iris biometrics, and most recently, general object detection and scene understanding. Savvides has generated more than 35 patents and patent publications, and over 50 unpublished patent applications to date.
What type of research will CMU/Cylab be doing for Oosto?
Our collaboration is all about helping Oosto penetrate new markets and business use cases with safety-related challenges. This means going beyond facial recognition. This includes developing reliable algorithms for object recognition, weapons detection and behavior recognition (if someone slips and falls or shows active aggression).
There are a myriad of use cases where this type of recognition adds a ton of value for business customers looking to eke out more intelligence and insights from their existing video surveillance infrastructure.
Can you describe some of these safety-related use cases?
There are so many, but here are a few that we’re currently exploring. In construction spaces, we can use Visual AI to determine if someone is not wearing a safety helmet. At CMU’s Cylab, we are developing AI algorithms to detect if someone has a weapon or, for example, has fallen down.
We’re even looking at AI to detect fighting or aggressive behavior in real-time. In retail environments, we’re exploring how visual AI can detect shoplifting or other kinds of theft use cases.
Are these real commercial use cases that organizations are asking for?
Absolutely. In today’s economy, many commercial customers are trying to do more with less. They want to stretch their dollars and investments from their current camera and video surveillance systems.
Thanks to advances in edge computing, we can now extract so much more operational intelligence and insights from video footage than we could even a year ago.
What types of industries are looking for this intelligence?
All types. From schools to commercial buildings, from casinos to construction, and public transportation – they all have a need to protect their people, property and intellectual assets.
Why is Oosto pursuing these safety-related use cases now?
Object and behavior recognition is the next step in Oosto’s evolution. Oosto is already the leader in real-time facial recognition – a technology which is very difficult to develop. But, we want to expand beyond this core technology and push the envelope in new verticals.
Detecting weapons or when a person has fallen down are just two examples that represent a huge potential to identify significant physical, operational and legal risks to modern enterprises. Oosto wants to protect the entire organization and this means not only addressing security risks, but risks that threaten the safety of its customers, guests and employees.
What excites you about working with Oosto?
Oh, boy. That’s a bit of a loaded question.
I’ve also been impressed with the strategic direction and humility of Oosto’s executive team, led by its CEO, Avi Golan. I know that’s an odd combination of traits, but research is all about making small, incremental discoveries – it takes time and patience to create true innovation. Oosto gets this and that’s what’s so exciting. We’re on the same page and are 100% aligned on what we’re trying to accomplish together.
On a personal level, I’m excited to work with Dieter Joecker, Oosto’s CTO, who spent 10 years as Chief Technology Officer of video security for Bosch Security Systems. CMU’s core expertise is developing AI-based IP, new patents, and algorithms, and we’re excited to collaborate with Oosto to build out their patent portfolio.
Another area that gets me excited and my blood flowing is how we can leverage edge computing and bring more of the compute power to video cameras or near-edge compute devices. We’re looking to exploit the power of edge computing to magnify the impact of our AI-based technologies and help generate more insights from the same video streams.
How exactly are you working with Oosto’s own development & AI/ML teams?
My team at CMU is strictly focused on behavior and object-based recognition. We’re collaborating with Oosto’s Belfast team (who are largely dedicated to facial and body recognition) to effectively productize our AI into working solutions for Oosto. This part really excites my team as they get the unique opportunity to see how their ideas and algorithms can be transformed into working solutions that help solve some of today’s most challenging business problems.
Which students are part of this effort?
This collaboration includes a mixture of ECE (electrical & computer engineering) PhD candidates, Masters students, and research staff at our Biometrics Center based in Pittsburgh.
Did you consider collaborating with any other facial recognition vendor? Be honest.
Yes. We did.
In fact, we were approached by other leading players in the visual AI space, but they lacked the market maturity, humility, and understanding of how to collaborate with a university think tank. Oosto already has a highly developed research team in place and a very mature engineering team who has experience operationalizing AI into working products for industry.
It’s also imperative that there is rapport and a shared enthusiasm for our research charter. I have genuinely enjoyed working with Avi, Dieter, and Oosto’s executive crew to define our mission and ensure that it connects with Oosto’s mission for making the world a safer place.
This shared vision is critical to our research team’s success. I’m confident that my research team will have the opportunity to apply our AI and deep learning algorithms to help realize, in part, this lofty objective.
