The Evolving Intelligence of Analytics Systems
In a recent article in the NY Times (Facial Scanning Is Making Gains In Surveillance - 21st August 2013), eminent journalist Charlie Savage has pointed out that the US government has spent millions of dollars on Facial Recognition technology for Crowded Scenes with disappointing results. However, this was an article about the experiences of a US government department with one US supplier of a single technology. It did not attempt to comment on the fact that governments in other parts of the world are already using iOmniscient’s Face Recognition in a Crowd technology very effectively in real world situations in combination with many other iOmniscient technologies.
This is not an unusual case. The Australian government recently funded a massive research project to evolve a traffic management capability – only to discover that the much more advanced iOmniscient technology had already deployed for several years.
These cases are not surprising as universities and government bodies do not have the visibility on the research done by private corporations.
The technology is moving rapidly in several new directions. Beyond security the technology is used for safety and the increase in operational efficiency. For the first decade, we were product centric. We have become use case / application centric. We felt we should share how it has evolved and where it has reached.
Legacy of the Past Decade
When we first started out, we differentiated ourselves by focusing on algorithms for coping with crowded scenes. A significant part of our patent portfolio is focused on the problems of crowded and complex scenes. No one else can find an abandoned object in a crowd. And our obsession with addressing the problems of crowded scenes has extended to counting, crowd management and finally to Face Recognition in a Crowd.
And we developed NAMS – an artificial intelligence based capability for reducing false alarms which was the most significant problem faced by CCTV systems then and now.
The behavior analytics evolved to address more complex behaviours such as counting in a crowd, queue management and the detection of slips and falls and smoke and fire.
The most significant advance was the development of recognition technologies for License Plates and for recognizing people in crowded uncontrolled environments.
Today is Different
Of course, all these algorithms keep improving as part of their natural evolution. But this is still a very product oriented approach. Even though we can cope with crowded, complex and realistic scenes providing end to end video analytics, from the point when an event is detected to recognizing the culprit, we have realized that the customer needs more than that.
But the next phase is with us. We have already moved beyond video based analytics to also using other senses such as sound and smell. Each of these technologies is useful in its own right.
The first breakthrough was to go beyond individual analytics capabilities to being able to address the requirements of particular Use Cases. So the focus moved from selling individual products to solving particular problems as described in a Use Case.
An example will explain this. If a person just falls down he could have slipped. If at the same time we hear a rising sound of people shouting and a crowd gathers, it is likely to be a fight. If a gunshot is heard, we can surmise that the person may have been shot.
Face recognition can then be activated automatically to attempt to recognize those that are present at the event. This ability to combine the information from complex sequences of events, both from real-time video and forensic footage, to draw a conclusion is an important advance.
Enhancing this focus on how our customers use the technology rather than just on the technology has led us to put together 30 Industry Packs focused on 30 different industries. These packs use the standard technology building blocks that the company has already put in place to generate solutions for specific problems faced by these different industries. The advantage of this user focused approach has allowed us to work in environments with multiple stakeholders such as in Smart City multiple stakeholders such as in Smart City deployments where multiple government agencies have to work co-operatively.