Results are in for our 2 poll questions, "Do Video Analytics Work?" and "What, if any, improvements would you like for Video Analytics?".
Do Analytics Work?
- 64% of respondents said, "Yes," video analytics does work
- 36% responded "No."
Improvements Wanted for Analytics
The breakdown for improvements selected by respondents:
- 55% - Better Performance
- 26% - Easier Setup
- 12% - Lower Cost
- 5% - Other
- 2% - No Improvements Necessary
176 people answered the polls. They are obviously non-scientific. Specifically, respondents self-select whether to participate so people with agendas are more incited to vote (video analytic vendors, etc.)
I abstained from voting. However, I would have voted "Yes" given the choice between the two exteme options. Like many, to me, the issue is how well analytics work, in what conditions and with what setup.
What Does the Responses Mean?
It seems clear significant concerns exist about analytic's performance - both from the 55% who choose better performance as their top need and the 36% who said analytics did not work (separate polls).
While I am skeptical of polls (mine included), the results are in line with what I hear continously. Here are 3 recent on-line examples:
- SDN Editor: "Video Analytics went from being considered the holy grail of technology to end users experiencing that it was not working in real-world applications." (August 2009 on-line edition [link no longer available])
- ObjectVideo VP of Sales: "I am tired of hearing [Video analytics don't work]" (August 6, 2009 blog post [link no longer available])
- ADT Application Manager: "Everytime if I talk about Video Content Analysis most people try to convince me by the fact that it won’t work!" (August 4, 2009 blog post)
These are just the most recent ones where I could provide a link/public reference.
What to Do?
I would recommend video analytic vendors provide detailed technological and operational information including the specific setup and optimizations required and the various issues that impact performance, test results with videos and graphs, etc. - and make it ALL public. [A good example is from an integrator on deploying ioimage for perimeter protection
So much of the expectation problem and the frustration that you hear from industry people comes from unclarity about what analytics can do and how it can be made to do. Detailed, public information would be key in solving this.
At the same time, I worry that this recommendation is naive. This seems to be an obvious thing to do but it goes against the history of the video analytic market. However, given the increased dissapointment in the analytics market over the last few years, maybe it can happen now.