Scylla AI Video Analytics Company Profile

By Sean Patton, Published Aug 29, 2019, 10:57am EDT

Scylla, an AI analytics startup, says they are targeting 1 Billion dollar valuation in 5 years and it "is not rocket science" to detect weapons and threatening behaviors with AI video analytics.

** ***** **** ****** to ****** ********** ***** offerings, **-**-******, *** *** they **** ********* *** trained ***** *********.

** **** **** ** look **:

  • *** ** ******
  • **** **** ** ***** Analytics
  • **** ********* ** ****** Targeting
  • *** **** **** ** Cost
  • **** ** ******'* **-**-******
  • *** *** ***** ********* Trained
  • *** ** **** ******** False *********

Scylla ********

******* ** ****, ****** was **** *** ** an "********** ***",**********, ***** *** ~** employees ******** ** ** platform *********** *** ********. Scylla *** ********** **** both ******* ***** ****** *********, *** ******* *********** software *********** *** ********* startup **********, ** ******** to *** * ***** as * ******* ********* for *** ******** ******** of *******. ****** **** they **** ** *********, in *** **, *******, and *******, ********* ******* on ******* ***********, **** multiple ********* ******* *********** in **** ******* *** machine ********.

****** **** **** **** initially ****** ** *********,****************** ****-***** ********, **** * ~$***,*** investment, *** **** ****** a ****** *********** **** of $* *******. **** said **** **** ****** half ** *** ****** round ** *******. **** are ****** *** * billion-dollar *********, ***** ** $100 ******* ** ******* by ****-** ***** ********* to ** ** ***** for $***,*** ** ******* from **+ ********* ** 2019.

***** *** **** ******* is ********* ** ****** to ** **** ***** $1 *******, ** **** be * *********** ********* to ***** $*** ******* in *** **** * years.

Not ********** - ***** ***** ********

***** ****** ********* ****** "we *** ** ******* active ******** ******", **** said ***** ******* ******* do *** ******* ********* but ***** ** ******** the ***** *** ****** for ***** **********, ***** would **** *****.

**** *-****** **** ******** Scylla's ***** *** ***** real-time ****** *** ******* behavior ********* *******:

****** **** **** *** working **** ****** *** behavior *********** *** ********/******** risk *********** ** ******* a ****** ** "******** the ****** ***-****** ********* and ******* ** ***** alarm". ***** * ****** like **** ***** ** a *********** ************ *** actually ********** ** ****** shooting *****, ****** **** "it ** **** ******" and ******** *********** ********** development *** ********.

Real **** ****** *********, ******* **************, **** ***********

******'* ****** ****** ****-**** preventative ****** ********* (***), ******* ************** *** tracking (***) ********* ********/**** ***********. ****** **** ****** detection ** * *********** of ****** ********* (*.*. knives, ****) *** ******** pattern ********. ****** **** supports ********* ****** *** vehicle ******** ******* ******** cameras, ***** ****** **** does *** ******* ********* a ****. ****** ********.**% ******** **** ******** people *** *************** *******. ** **** few ********* ***** ** offer **** ***** ** cross-camera ******** ******** ** production *** **** *** not ****** ******, ** are ********* ** **** figure.

**** ***** ***** *** Scylla ********* **** *********** scores *** ****** *** violent ********:

****** **** **** ***** vehicle ************** *** **** recognition *** ********* ********* features ** ***** ******, noting **** **** ** not ***** ************* ****** vehicle ************** ** **** recognition *********** **** ***** manufacturers.

**** **** **** ******* were **** ********** ** face *********** *** ******** to ******* ******* ****** the ********, ************ ** close ********* ** ********.

****** **** **** ** not ***** *** *****, images ** ****** ** default, *** **** **** store **** ** ********* by *** ***-****.

"Millions ** ******" *** ********

* ******* *************** ****** noted *** ******** ***** neural ******** ***** ******** of ******, ***** ****, prior ** ***** **** out ** **********. **** said **** ***** ****** image ******** ******** ** public *** ******* ******. These ***** ** ****** are ****** *** ** developers ** ******* ****** image **** *** ****** available, *** ********* *** quality *** ***** ************ training. ****** **** ***** datasets **** **** ******* over **** ***** ** AI *********** ** ********** and ******* ***** ******* to ** *** ** production ******* *** ******** retraining ***** *** ********'* cameras, ****** *** **** recognition.

**** **** **** ***** 2+ ***** ********** *** updating ***** ****** ** decrease *****-******** ***** *** smartphones, ******, ***** *******, etc.

************, **** **** **** trained ***** ****** ***** thermal *** ********* ** black-and-white ******* ***** ** use **** ********* ******* cameras ** ****** ********/******* deployments.

VMS ******* ******* ** *********

****** **** **** **** have**** *** *********** **** Milestone, *** ***** *** supported ** ********** **** streams ****** **** ******* or ***** *****. ********* systems *** **** ****** alerts ****** *** ******** Smart ******, ** *** the ****** *** ****** and ****** ***. ***** with **** ** ***** VMSes *** **** **** to ******* ****** **** the ****** *** *** mobile ***. **** **** they *** ********** **** deep *** ************ ** 2019.

*** ****** ****** *** alert ***** ** ***** of *** *******, **** a ********** *****/******** *****, in ******** ** *** links *** ********** ** Scylla's ********/**** *********** ******.

Using ********** **** ****** ****** ********

****** **** **** *** not ******* ***** *** neural ********, *** ******* use ****-****** ********* (*.*. YOLOv3, ****** *-***). **** said ***** ********* *** able ** ****** * weapon ****** ***** ** being **** ** ******, at ~**'/*** *** ~*****. They ***** ***** ********/******** detection ** ******* *** behavior ******** *** ** not ***** **** *********, claiming **** *** *** resources *** ************* ****** for ****.

******* *** **** ***** analytics, **** ***** **** sufficient ***** ******* *** good ******** *** *** biggest ********** ** *********.

On-Premise ********* ** ******

****** **** **** ********* performing ***** ********* **-*******, with ****** *** ******** to ******* *** *********. They ** *** ********* offer ******** *** ***** that **** *-**** ********* they **** ********* ***** hardware *****.

**** **** **** *** NVIDIA ******** ************ *** dependent ** *** ********* analytics ** ***, *** noted *~$*,*** *** **** ****** ******* ** *******. They **** *** ****** with *********** **** ******* the ***** ******** ***** on ******* ****** ******, automatically ************ ******* ********* weapons *** ******* ********.

**** **** **** ***** a ****** ********, *** primarily **** *** *******, that ***** ******* ********** system ***** **** *** many ******* ** *** a ***** ********.

*******

****** **** *** **-******* systems, ******* ** ***+ camera *******, ** **********, education, ********, ****** *******, the ******* ** ****** based ** *** ********* required *** ****** ** the ********, *** ********* cost $** - $** per ****** *** ****. Under *** ****** ******* would ******* $*** *** camera *** ****. ***** is ** ********* ********* option, *** ****** **** many ********* *** ********* multi-year *********.

****** **** ***** ***** monthly ************ ***** $** per ****** *** ***** and ******** *** * analytics.

Channel *********

******'* **-**-****** ** ********* with ******* ********, ********** outside *** **, *** they **** **** ****** to *** ***** ***** on ********** ******** ************. They **** **** **** significant ******** ********** ********* and ********** **** ********* AI ********* ********* *** the **** * ***** at **********.

Comments (4)

Interesting, second gun analytic startup(ZeroEyes) in last couple month with YoloV3 .

Scylla said they did not develop their own neural networks, and instead use open-source solutions (e.g. YOLOv3, Faster R-CNN).

Curious, why would one need both of those networks? Faster R-CNN and Yolov3 do the same thing basically while Yolo v3 seems to faster and more precise by tests results. Object detection: speed and accuracy comparison (Faster R-CNN, R-FCN, SSD, FPN, RetinaNet and YOLOv3)

They said their analytics are able to detect a weapon within 400ms of being seen on camera, at ~40'/12m and ~10PPF.

So foot size gun would be 10x10 pixels and it is much stronger claim in comparison to ZeroEyes ZeroEyes Gun Detection Startup claimed more modest number: "Our current data shows we need roughly 17PPF/55PPM".

They noted their behavior/violence detection is looking for behavior patterns but is not using pose detection, claiming that the GPU resources are significantly higher for pose.

Would be nice to get more info on behaviour patterns analytics. This is very cool by itself ! Being cheaper (resource wise) then pose estimator makes it especially interesting.

Agree
Disagree
Informative
Unhelpful
Funny

Scylla's valuation of $1B in 4-5 years on $100M in revenue is right there in crazy-land with AnyVision's $1B revenue pipe dream by 2022. These ISV companies have grossly overestimated the revenue potential video analytics as an ancillary "feature" of what will become table stakes for the camera companies in 5 years.

Agree: 3
Disagree
Informative
Unhelpful
Funny: 1

as an ancillary "feature" of what will become table stakes for the camera companies in 5 years

I generally agree with this statement. To play devil's advocate, though, look at what is happening in China - analytic focused startups selling large AI systems directly to governments, generating already hundred plus million revenue per year (e.g., Megvii Financials and Growth Examined).

Of course, these Western firms have virtually no chance inside China and the non-China market for this is likely much smaller / much less able to sell larger scale analytics to governments.

Agree
Disagree
Informative
Unhelpful
Funny

We give for standard assumption that all the start-ups always form with the goal to grow and to capitalize their investments by selling their products/service/competence.. The question is, which is the product? Claims can be a product themselves in a certain sense, don’t they?... I mean I actually see more and more often, more financial speculations than industrial projects... Not necessarily meaning about this case, of course..

And of course, I am writing just driven by envy...πŸ™„πŸ˜…

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

This article is part of IPVM's 7,201 reports and 959 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