Let me turn the questions to you.
Why would I use a service for people counting, intrusion detection, etc., if my camera already included that for free? That is a fairly common offering.
How are you differentiating against existing analytics? What are your competitive advantages?
If you defined that, I think you would get more people interested in doing VAaaS.
Pro Focus LLC | 04/05/16 01:07pm
I think the only way a VaaS could differentiate itself from built in analytics is in the reporting and data gathering. If you could find a super easy way to generate reports and present them to end users, you may have something. Otherwise, analyzing video data only is pretty common.
I think this would have to be marketing based info for people to pay for it. I don't have any clients that would pay for this for security purposes. Not to say there aren't any, I just don't know of them.
That said, I don't know that this isn't already a service offered to retail markets.
All of these new "Service as a Service" things are just buzz words used in marketing presentations where the word "leverage" is commonly used in tandem.
"Leverage the Service as a Service to drive the execution of your strategic blah blah"
It's a fancy word for subscription-based service. Marketing people have determined people don't like the word "subscription" so they've changed it to "X as a Service".
For that reason alone, I'm out.
For others considering Service as a Service in anything, reword the proposal to yourself using the word "Subscription" or "Service Contract" and see if you still like the shiny marketing materials.
I'm in a crabby mood today... no offense intended to anyone.
I think there is one potential use case where it could make sense.
The CPUs in cameras these days are pretty weedy things for doing edge based analytics. As all of the encoding has moved from DSPs into dedicated logic it has left little in the way of general purpose processing cycles for analytics. So we haven't really seen any improvement in the market for some time. Nobody is doing anything smarter than intrusion detection and people counting.
As Google and Deepmind and all those other guys are demonstrating, if you have lots of CPU cycles (and maybe some GPU cycles) you can do some clever stuff. Have you every tried a search through your personal Google photos? It's pretty amazing.
So if you can get your video into the cloud, then I think you can do much better analytics than we see today bundled with cameras. But of course you then have a bandwidth problem streaming all that video out into the ether.
So perhaps there is an opportunity for a mixed-mode approach. Where the basic analytics is done at the camera end to reduce the video down to a scene description or some other metadata, which is then streamed to the cloud for further processing. The kind of applications you might be able to offer with such an approach could be much more accurate video analytics for example, where the system is constantly learning and improving over time to reduce false alerts to levels which are not possible today. Another could be more accurate people counting (where now the only option is a stereo, thermal or time-of-flight sensor).