Deep Learning Surveillance Startups Deep Problem

By John Honovich, Published Jun 23, 2017, 08:07am EDT

The undeniably good news for the video surveillance market is that we are seeing the rise of more startups than in many years.

The cause of this is clear - deep learning / 'artificial intelligence' / the next wave of video analytics. While video analytics in video surveillance had its initial entry almost 15 years ago, this one promises to deliver far higher accuracy and intelligence.

But even if the technology delivers, a critical business problem remains for these startups.

Problem ** ***

*** ******* ** ***** these ******** **** *** in *** ********. *** will **** **** ***** with ** ******* ********* players ** *** ******? The ****** *** ******** hardware ******** ******* **** advanced ***** ********* *** are **** ** ********** low ******. ** *** opposite ****, ***** ******** generally **** ** ********* or ********** ** ***** surveillance ********. **** ** they **?

'Add **' **** ********

*** ******* ***** ** see ** **** ***** surveillance ******* **** ******** startups ** ** ******* an '*** **' *******, such **** *** **** brings ***** *** ******* and ***** *** ********* and *** **** ******** providers ***** **** ***** video ***** *** ********.

*** ***** ******** ** this, *** *** *******, is *** **********. *** startup ******* ** **** it ****** **** **** (developing ********) ***** ** avoids ****** ** ******** well-established ****** ** ********* hardware.

Cost *** ********** ** '*** **' **** ********

***** ************ **** ******** startups ********* **** ** sell ***** ******** ***, in **** ***, ********* as * ******* (** maximize *********** **********). **** creates * ********:

  • * *********** ******** ** cost *** *** *** user: *** *** **** still *** ** *** everything **** *** **** needs ** *** *** this *******. *** ********** case ** **** **** service ******** **** **** offsetting ***** *** *** risk ** **** ******** incumbents ***** ***** *** rival **** ******** ** their ******** *********.
  • ********, ****** ** * 3rd ***** **** ******** analytic ******* ******** *********** with *** ******** *********, which *** ******* ************* issues, ********* ******** *** monitoring *********** (*.*., ********** the ********* ******** **** the *** / *** main ********).

Incumbent ***** ***********

********* ****** (*** ******* ************* *** already ******* ****, *** ** ** clear ****** *** ********** will *** ********* **** time) *** **** ******** / ** ** *** next ******* ** *** to ***** ******** *********, as * *** ** differentiate ********** *** ***** back ******* *** ******* cycle ** ******** ***** cuts **** *** **** few *****.

** *** ****** **** incumbents *********** **** ******** into ***** *** ******, recorder ******** ** *** software, **** **** ****** certainly **** *** * significantly ***** ***** **** and **** ******** ******* complexity **** ******** *** are ********* ***-** *********.

Hope *** ********?

** *** * ******** paths *** ********:

  • ***** ***'* *** ******** and *********** ***** *** deep ******** ********* **** those ******* ** ********* such **** **** *** complexity *** ** ******* (e.g, ** ***,****** *** ******* ****, though **** ********* *** challenges ** ****** * competitive ****** **** ******* mature *********). ** **** way, ******* **** **** learning ******** *** ***** out ***** *** ******** line **** ****** **********, differentiating ** ***** ******** analytics. ** **********, *** it ***** *** ** simple, **** ***** ******** their ********* ** **** incorporate ******** ***** *** target * ****** ******* of **** ********'* *******.
  • ***** **** *** *******'* deep ******** ** *** superior ** **********, ******* having * **** ****** and ******* ******* ********. Convince ** ********* **** buying *** *** ******* and ************* *** *******'* software **** ** *********'* existing **** **** ********** the *********. **** ** easier ** ********** **** path * ***** *** almost ********* **** ****** in * *** ***** valuation.

******* ** ***** * paths, ** *** ******** analytics **** '*****' *** without ********** * *********** performance ******* **** **********, many ** **** ** these ********* *** ** risk ** ******* ***.

Comments (15)

John:  Interesting take.  Q: what ever happened to BRS Labs?

Agree
Disagree
Informative
Unhelpful
Funny

Robert, BRS Labs sort of exists as twice rebranded / reformed entity (e.g., BRS Labs Is Gone; Giant Gray Is Here) and with more drama (i.e., BRS Sues Founder, Allegedly Stole $15 Million).

An optimistic view on BRS Labs is that they were too early. A pessimistic view on BRS Labs was that they were more interested in hype and fund raising than execution.

Agree: 5
Disagree
Informative: 1
Unhelpful
Funny

Has anyone here worked with Giant Gray? I recently spoke with another a colleague in manufacturing and he believes they have promise, but I don't know anyone else who has ever dealt with them. 

Agree
Disagree
Informative
Unhelpful
Funny

Has anyone here worked with Giant Gray? I recently spoke with another a colleague in manufacturing and he believes they have promise

Many people worked with BRS Labs (subsequently Giant Gray, now Omni Ai) over the years, some were initially positive, essentially nothing has worked out.

Given all the damage they suffered, suing the company's founder, etc., I find it hard to understand why one would take a risk with them at this point.

There's a number of emerging deep learning, abnormality detecting companies developing (e.g., see our deep learning list and in particular, iCetana which claims similar approach).

Agree: 1
Disagree
Informative: 2
Unhelpful
Funny

Thanks for the reply. That is pretty much were my mind was at.

Agree
Disagree
Informative
Unhelpful
Funny

John - Maybe another goal/hope, for those new start-ups, is to be acquired by the traditional video manufacturers. Cash out big. A lot of traditional video manufacturers are falling behind in term of VI/AI

Agree: 3
Disagree
Informative
Unhelpful: 1
Funny

Maybe another goal/hope, for those new start-ups, is to be acquired by the traditional video manufacturers.

For sure, that's what I describe as path #2 at the end of the post.

Cash out big.

I am skeptical about that cashing out big because that assumes the startup can get great traction (which is hard as an add-on) to justify that valuation or that the startup's technology is just so much more advanced, and the incumbent views it as such, that they are willing to pay big (which historically has not been easy to do).

Agree: 2
Disagree
Informative: 3
Unhelpful
Funny

Thanks for the info and links John.  

Wow!  Another Bernie?!

Agree
Disagree
Informative
Unhelpful
Funny: 1

Thanks John, very interesting article.

Agree
Disagree
Informative
Unhelpful
Funny

I'm curious.... do the various start ups own any patents on their deep learning technology?

Can an algorithm be patented?  Algorithms are largely giant and extremely complex 'if/then' computations, no?  How is that patentable?

And if the deep learning technology is not patentable, then how can any one provider 'cash out big' if any of the industries existing players can just develop their own deep learning algorithms?

Agree
Disagree
Informative: 1
Unhelpful
Funny

FD I'm CEO of Camio. This perspective is super-helpful for understanding how to approach the industry.

But the surveillance industry today feels like PCs in the early 90s when people viewed PCs as tools for spreadsheets and word processing.

Deep Learning itself isn't really a feature so much as an enabling tech, so the paths for startups are predicated on what we enable people to *do* with the video and audio. The two paths mentioned (to become either a) competing vertically integrated hardware stack or b) feature of incumbents' product line) are secondary to the growth we could create by making video surveillance become the eyes and ears for machines that learn. That requires collaboration - similar to the way Web 2.0 APIs and layered services created businesses much bigger than standalone PC productivity tools.

Agree: 1
Disagree
Informative
Unhelpful
Funny

That requires collaboration - similar to the way Web 2.0 APIs and layered services created businesses much bigger than standalone PC productivity tools.

Examples please of that?

Are you saying that these new markets are going to be much bigger than existing markets for video surveillance? If so, please elaborate with examples.

Agree
Disagree
Informative
Unhelpful
Funny

Yes, as one example the $190B fishing industry has a tragedy of the commons economic problem. They lose $32B to illegal and unreported fishing. When computers can solve at low cost via Electronic Monitoring, they unlock $32B in one industry alone.

Agree: 1
Disagree
Informative
Unhelpful
Funny

When computers can solve at low cost via Electronic Monitoring, they unlock $32B in one industry alone.

But what's the potential revenue / market size for deep learning video providers? If it's $32B congrats, you'll be buying Tyco in a few years. But I am assuming the actually revenue potential is far far lower. 

Second, why can't incumbent companies with existing hardware (take Hikvision since they have cameras are aggressively targeting deep learning) take that business?

Agree
Disagree
Informative
Unhelpful
Funny

1) yes, the revenue to tech providers will be far less than $32B saved annually and the revenue will go to providers across the whole stack from integrators to cameras to NVRs to communications to deep learning labelers to data analysis to human review etc... meant to cite it just because it's an example of a market opportunity that can't be solved simply by adding a DL feature to a camera.

 

2) same reason Microsoft didn't take the Web. The solutions require more than one company's expertise.

Agree
Disagree
Informative
Unhelpful
Funny
Read this IPVM report for free.

This article is part of IPVM's 7,338 reports and 973 tests and is only available to subscribers. To get a one-time preview of our work, enter your work email to access the full article.

Already a subscriber? Login here | Join now
Loading Related Reports