Why Video Analytics As We Know It Is Going The Way Of the Dodo
At my SecurityDreamer event at the Hard Rock Cafe in Atlanta a few years ago I discussed the idea of the difference between video analytics one-dot-O (1.0) and video analytics 2.0
The idea I wanted to get across was that popular video analytics vendors at the time were already appearing stuck in processor-intensive analysis of pixel changes. Video analytics products that were fashionable then, such as Object Video, ioimage, Cernium, Vidient and others had products that were limited by the specific algorithms they used, the types of cameras or images they supported (day/night/indoor/outdoor/scope/frame rate/etc), and the amount of configuration and tweaking once deployed. Unless a video analytics product was perfectly matched with the optimal camera, environment and algorithm, the results were always disappointing.
That description of video analytics sounds very similar to the way the first computer-based access control products of the 1980s were described. It is similar to the first firewalls and IT-intrusion detection systems. It sounds just like the first Microsoft Windows operating systems. In other words, It had all the characteristics of a “version 1.0” technology.
1.0 is always exciting. 1.0 introduces new ways of solving problems, opens new avenues and gets our creative juices flowing. 1.0 gets us turned on and many of us want to use the technology because we are fearless or desperate for some solution. But 1.0 always comes with a price of inefficiency and ineffectiveness. 1.0 always makes us long for a new version that will achieve the desired end less painfully and more effectively.
2.0, in an operating system, IT software, or any technology generally represents a quantum leap forward. A rethinking of the problem and an entirely new approach. Consider the breathtaking difference between Microsoft DOS 3.3 and Windows 95. It was like the PC had been reinvented. Suddenly processes that took dozens of configuration steps were automatic and graphically appealing. DOS 3.3 became obsolete.
Video analytics 2.0 had that same promise – the attractive idea that the headaches and misfires of video analytics 1.0 would be replaced by an automatic, intuitive system. The company that seemed to have a true video analytics 2.0 approach at the time was BRS Labs – A startup in Texas that had yet to be proven. The BRS difference was its use of machine learning. An analytics system that was self learning seemed to be exactly what the end users were hoping video analytics 1.0 should have provided. Those who were not exposed to BRS scratched and clawed their way through video analytcs 1.0 deployments, steadily reducing their expectations. Today Video analytics 1.0 vendors have happy customers – customers who expect little and receive what they expect.
Fast forward to today and we see Vidient out of business; Object Video is no longer a technology competitor and has instead reoriented it’s business to patent protection; other analytics vendors are selling very specialized solutions. And then there is BRS Labs – making money, growing every quarter, and silently showing the world what video analytics 2.0 truly represents. As more prospective customers see what video analytics 2.0 represents, the 1.0 vendors will drop like flies.
Video Analytics technology has struggled over the past decade with products that require expensive and time consuming installations, high ongoing maintenance costs, and unacceptably high false alarm rates and overall poor operational performance. In short, traditional video analytics technology has not proven viable and the market, as all markets do, requires innovation. The market craves an operationally effective solution and innovation is the key to satisfy market demands. As I have said since 2008, the learning approach taken by BRS Labs is ushering in video analytics 2.0. The broader market is now coming to this same realization.