AWS Panorama Examined

By Isabella Cheng, Published Dec 08, 2020, 11:07am EST (Info+)

Amazon recently announced the AWS Panorama, a machine learning appliance that is designed to turn non-smart IP cameras into smart cameras.

IPVM Image

In this report, IPVM examines:

  • How AWS Panorama works
  • Advertised pre-built CV applications for niche use cases
  • CV at the edge
  • AWS Panorama Device SDK
  • Panorama partners
  • Pricing
  • Opportunities and risks

How ** *****

*** ********, ********* ** ****, ** described ** * ****-**** **** ****, box-shaped ********* ****, **** ********* ** power, *** ** ********** ** *** AWS ********** *******, **** ************* ******** ONVIF ******* ******* ** *** **** network ** *** *********, **** ****** machine ******** (**) ****** ** ***** camera ***** *** ***** ******** *********.

**** ***********, ******** ** *** ** the******** ******* ****, ***** *** ******** ***** ** three *****.

IPVM Image

*** **** *** ********* "******** ********** to ******** ****** ******* ** * given **** *** ******** ******* ******** ML ****** *** ******," *******, *** not ******* ** ***** ** * maximum ****** ** ****** ******* * single ********* *** ******* **.

Augments ***/*** *******, ****** ************ *********

** ******* ****** ** *********** **** to ******* ******** ***/*** *******, ** be **** **** **** **** ***** offerings, *** ** *** ***** *** analytics. *** *******, **** ********, * customer (** *** **********) *** ***** their *** ******** ****** ************, ** source **** **** *** ********, ******** a **** ***** ** *** ***** while ******** **** *** ******* ***. The ********* *** ******* ******* ***** available ** ************, ** ** ** able ** ***** **** ***** ********* on *** *** **** *** ** not ******** ** * ***/*** ***********.

Niche ******** **** ***

***** ******** **** ** ***** * variety ** *********, ***** ********* **** likely ** **** ********** ** ***** who **** *********** ************ *** ***** situations. *** ******** ****** "***-***** ************ from *** *** *****-***** (**) **********," and *** ************ *** ************ ******* out ** *** *** ** *** product **** *** ************ ********** ** manufacturing ** ********** **********, ********* ************* quality *******, ****** ********, **** ******, and ****** ***** **********. ***** ************ are *****, *** ************ ************* *** unlikely ** **** *****-**** ***-***-***** *********. In **** ********, ***** ***** **** with *** *** *** ******** ** specifically ****** * *********** ******** *****.

*******, **** ********* ******** *** ************* come ** *** **** ** **** greater ********** *** ****** ***** ***** than ** *** ******** *** ****** make *** *** ** *** ******** analytics ** ***** ******** ***** ************ products. *** **** ***** ***** ***-***** applications **** *** **** ******* ***, but **** ***** ******* **** ***** analytics ***** **** ***** ******** ********, or *** ****** ******** **** ***-***-***** from *************. ********'* ******* **** ****** that ***** ***-***** ************ ******* "****** counting, ******* **************, ******* ***** ***********, PPE *********, *** ****** ***** ******." Unlike *** **** ***** ********** ************ above, **** ************ ************* ******* ***** these ******* ********* ***-***-***** ** ** specialized *******, **** ** *** *******. Users **** ****** *** *** ** use ******** ** ********** ***** ****** tasks.

*******

********'**************** ******* ***** *** *** ********* ($4,000), ******* *** ****** **** ($*.**) for ***** *** *********, *** ***** storage *****. *** ******** ********* ********* kit ** **** ********* *** * one-time ****, *** ** ********, ********* on ******* *** **** ***** ** update ***** ** ****** *** ******** logic. *** ******* ********* ** ** follows:

  • *** ******** *********: $*,***/******
  • *** ******** ********* ********* ***: $*,***/******, however, **** *** ********* *** ** necessary ** ****** ******** ************
  • *** ******** ********* *****: $*.**/***** *** active ****** ******
  • ***** ******* (** ***** ** ****** and ******** ***** **** ** ******** Appliances): $*.**/** *** *****

IPVM Image

******** ** *****-** ****** *** ******** analytics, ***** **** ** **** ********* but **** ***** ** **** ********* than ***** ******** ******* **** ***** charge $** - $** *** *******, per *****.

AWS ******** ****** *** **** ** ***** ******, ********* ********-******* *******

***** **** *** ******** *********, *** also ******** *** ******** ****** *** for ****** ******** ** ***** ***** cameras ******* ** ********** ** ******** services ******* *** *********. *** ******** device ************* *** "***** *** **** appliances *** ***** ******* **** ****** more ********** ******** ****** ****** ** the ****" **** *** ***. ***** devices, ****** "********-******* *******", **** ******* APIs ****** ** *** ***** **********. AWS ******** *************** ******** ****** ****** ******* ****** *** Ambarella ** ** ******* **** ** be **** ** ********-******* *******.

Business ************* *** ****** ********: **** *** *******

**** *** ******* *** *** ************* who **** **** ***** ** ****** partners ********* ********* ********-******* *******. ***** CV *****, **** *** ******* ***** build *** ******* **** ******** ****** the ***** ** ********* ******** ****** without ***** ******** ** ***. *******, this *********** **** *** ***** ***** more ******** ************* ** ***** *** looking *** **** ***** ************ **** Panorama *** ******* *** **** *** Vivotek ** *** ******* *****. **** collaboration ***** **** ****** **** *** Vivotek's ***** ** *** ********* **** do *** ******* *** ****/******* ******** but ***** **** ** *********** ****** AWS ********.

Many ******** ********

*** *** ***** ** "******** ********" *** *** ******** ** *** application *** *********** ** *** ******** solution. ** ***** **** ** ******, integrating, ***********, *** ***** ********, *** has ****** *** ******** ******** ** its ******* *******. *** **** **** "the ******** **** ************ ***** ******* to ********* *** ********, ***** ********-********** CV ****** *** ************, *** ******** Panorama-enabled ******* **** **** ********** **** Panorama." *** ******** *** ****** **** four *** *****:

  • ****** *********** *** ********** ******** (***): Accenture, ********, *********, ***.
  • ******** ****** (**) ****** *** ************: Bigmate, ********, ******, ***.
  • ******** *** ************ ********: ********** ************, Stanley ********
  • ********-******* *******: **** **************, *******, *********, NVIDIA, ***.

***** *** ******** *** ********** ** specific ***** ** *** ******** ********, having *** **** ******** *** *** complexity *** *** ****. ********* ** a ****** ** ******** ** ** completed ******* **** ********* ******** *** add ****** ** ********* ***/** ********** that ***-***** ** *** ****. *******, end-users ***** **** ** *** ************ who *** *********** ********** **** ***** to ** **** ****** **** ******* in **** ********* ******** ********* ** the ***** ** ************ *** ******** solution.

Opportunities *** ***** *** *** ********

** *** ** ********** ** ********* Panorama ** ***** ********** *****, *** could ****** * ***** ********* ** cloud ** ********* *** ********* **** is ********* ** *** ********.

*********** **** **** ** *********** ** AWS *** ** ********* ** ***** another *** ********. *********** **** ***** needs *** **** ****** ******** ** fulfill ********* **** *** *** ********* ready-made **** * ************ ************.

*******, *** **** **** ** ******* with *********** ******** ********* *** *** offer * ******* *** ** ****** analytics ** **** ** ******* ** video ********** ********** **** ******* ****.

Vote / ****

Comments (13)

What are "ML Models"

And what would be a niche case example this could be used on. Im still slightly confused what the heck this is.

Agree: 5
Disagree
Informative
Unhelpful
Funny: 1

Hi Sean,

ML models mean Machine Learning models. Sorry I was not more clear about that. The article has been updated to reflect what ML stands for.

A niche example might be in an industrial application, for example, if an assembly line wants to train a camera to detect that a screw is missing in a specific piece of hardware that is coming off the assembly line. The idea is, AWS/its partners who specialize in making models can tailor one specific to your need to detect if a screw is missing from what's coming off the line. That is a more complicated/specific/niche analytic than, say, people counting or zone recognition, which basically every camera maker already offers off the shelf.

Let me know if that gives you a better sense of what Panorama is.

Agree: 2
Disagree
Informative: 3
Unhelpful: 1
Funny

Yes! Thanks very much!

Agree
Disagree
Informative
Unhelpful
Funny

This looks like a game changer... very promising.

Agree: 3
Disagree: 4
Informative
Unhelpful
Funny

Nice idea but too complicated and way too expensive

Agree: 7
Disagree: 1
Informative
Unhelpful
Funny

I'm wondering about your note regarding "way too expensive". Is that because you don't see value in this solution or are you comparing it to other options on the market?

Agree
Disagree
Informative
Unhelpful
Funny

$4000 for a device and monthly fee ? These type of devices are are already available for way less than that with no month fee

I just dont see how people will pay that

I understand these are niche applications but it seems like it can done cheaper already

Agree: 3
Disagree: 1
Informative
Unhelpful
Funny

Split responsibility is the problem I foresee. The end user still needs integration services for installs, but relies on Amazon for software issues? Or does Amazon sign up integrator partners for local support?

Agree
Disagree
Informative
Unhelpful
Funny

Hi UM3,

Amazon has signed up Stanley and Convergint for integration ("distribution and integration partners"). However, there is still the question of how well those integrators could handle these advanced software integrations.

Agree: 2
Disagree
Informative
Unhelpful
Funny: 2

They wont be able too. They will have to rely on Amazon or the developer of the application which they will upcharge for.

Agree: 1
Disagree
Informative
Unhelpful
Funny

What is IPVMs position on this pricing model? ;)

I foresee a lot of issues if people think they can just plug this in, and magically turn their old, poorly positioned, noisy and low-fi feeds into useful analytics . There are just too many caveats when setting up analytics which makes this a risky proposition.

Greenfield installs where there's a need for analytics might work, but I am not sure about the TCO vs just getting a better HW for a VMS with decent analytics (do they exists?).

Agree: 5
Disagree
Informative
Unhelpful
Funny

I see a couple of separate points to make here:

1) AWS is behaving like an ai-software aggregator for the market- but their value isn't ai, HW or even cloud.. it's pure market reach. The bad news for participating vendors is Amazon has reportedly used programs like these for market research to develop their own competing products.

2) Why introduce another point of failure (which happens)? AWS storage is only for the ML algorithm and settings. Nvidia has a similar program- but it's not cloud based. Link: Amazon details cause of AWS outage that hobbled thousands of online sites and services - GeekWire

3) Big Tech. It's concerning to have this data being logged in the AWS cloud. My concern would be that AWS wants your customers' data.

Example: Amazon's Ring reportedly collects product use data, but also collects which devices (iphone, andriod,) a user uses to interact with it, and potentially what other devices are on the same network. Hmm.

Now - put Panorama in that context: are you comfortable having your retail customers' "retail Insights" results running through AWS (the largest online retailer in the world) servers? I would tend not to go this route.

Links: Amazon's Ring logs every doorbell press and app action - BBC News - SamfordCrimson News

Links: Amazon faces backlash over using Sidewalk for neighborhood networks - Business Insider

Links: *privacy not included - Ring Video Doorbell

Agree: 4
Disagree
Informative: 3
Unhelpful
Funny

Tom Burns; very well thought out.

Amazon AWS "may"= Big brother. " Hi we would like to offer you this super great product with many features (some of these features being aƧcess to your "everything"). I imagine they gain access to your imagery and metadata outside of "for purely analytic application feature purposes" ( I guess "analytic access" would need to be defined). Has a creepy feeling to it AWS...

Agree: 1
Disagree
Informative
Unhelpful
Funny
Login to read this IPVM report.
Why do I need to log in?
IPVM conducts reporting, tutorials and software funded by subscriber's payments enabling us to offer the most independent, accurate and in-depth information.
Loading Related Reports