Startup Image Intelligence 10X Accuracy Guarantee

By Sean Patton, Published Mar 26, 2018, 12:56pm EDT

An Australian startup, Image Intelligence, is targeting the biggest problem of video analytics, poor accuracy, with quite a claim:

Is this legitimate or a stunt? What does Image Intelligence aim to do? We spoke with Image Intelligence's founder Jimmy Lee [link no longer available] to better understand their offering and the accuracy guarantee.

Company **********

***** ************ ** ********** this **********, ***** ** *** * employees ** ************* ******** ********* ** developers. 

*** ***** ************ **** was ********** **** *******.* *****-***** *********** ****** recording ******, **** *** in **** ** ***** on *** *********** ** the ****** ********* ********* they *********. ** *****, they **** ********** ******** with ***** ******-***** ********* alerts ***** **** ** high ***** ******* *****. This *** ******* ******-****-********* storage **** *** *****. To ******* *** *******, they ********* ******* ******** based ********* ** ******** people.

Only ****** / *** / ******* *********

***** ************ ****** ******, pet *** ******* *********, including **** ******. *** ***** below ***** * ********** score ** ****** ** vehicles **** ******* *** scene: 

**** ** *** ***** more ************ ********* **** as ********, ********* *********, people ********, ***.

Person ********

***** "*****" *********** ******* to **** * ****** ****** the ***** ***** ** ********* 3 ****** ** * person.

***** ************ **** ******:

* ***** ********** ** not **** ***** ** facial ***********, ** ***** clear **** ***** *** not ********* ** ******** camera *******. ***** ******** like ***, ****** *** skin **** *** **** in **** **********.

No ******** ***** / ** *** **** **

***** ************ **** *** record ***** *** **** it ***** ** *** user **. ***** ************ receives *****, ******* ******** people, **** ** ********, cataloges **** *** ***** it ********* *** *** to ****** ***** *******. ** API [**** ** ****** available] ** ********* ** integrate * ***** ****** by ******* ****** ********* generated ****** ** ***** to ***** ***** *******, and ******* * ********** score *** "******** ***" data **** **** *** cloud *******.

Offline ******* *********

** **-**** ******* ** the ******** ** ********* through * *** ******* to *** ** *** enabled ********, ***** ***** to *********** **** ***** cloud ********. ******* *** not ********* *** *** offline *******.

Pricing *** *****, **** ****

***** ************ ******* *** frame ******** ($*.** *** per **** ******), ****** most ******* ***** ****** per ******. **** ** a ****** ***** ** image ****-****** ********* *** much **** ********* **** Amazon ** ****** (****** ***********, $* *** **** frames),****** ***** ****** ** $*.** *** **** frames).

*******, *** **** ***** be $*.** *** ** hour ** ***** ** 10fps, ***** ***** ** incredibly *********. **** ** 1fps, ** ***** ** $0.36 *** ***** ****, or **** $*** *** camera *** *****.

3rd ***** ***********

***** ************ **** **** have ** *********** **** an ***** ******* ********* *****, *** *** *** clarify ***** *******. ************, they *** ******* ** other *** ************.

Guarantee ********

***** **** *** ******** a ******* ** * frame, **** *********** ****** you * ***** ($*.**) for **** ***** (*.*., 10x **** *** **** for **** ***** - 1/10th ** * ****). This ** *** ***** enough ** ** **** and ******** ****** *** of **** *****. ** you *** ****** ******* issues **** ********, **** any ***** ********, *** best ******** ** ** stop ***** ** ** the ********* **** *** 

Outlook *** *******

***** ***** ************ *** eye-catching *********, ***** ******** suffers **** * ****** of ********:

  • **** *******
  • ******* ********* (********, ********* violation, ****** ********, ***.)
  • **** ** ***** ****** web *********
  • **** ** *** ***** VMS ************

** ****** * ******* contender ** ***** ************, they **** **** ************* reexamine *** **** ******** the ******.

Comments (9)

However, the cost would be $36 for an hour of video at 10fps, which would be incredibly expensive. Even at 1fps, it would be $3.60 per video hour, which over a month would still be $1,000 per camera.

It is not clear exactly what the specifics of the 10x refund are, but this could be the next big thing after the crypto currency bubble.

"Guaranteed accuracy" would be very hard to maintain over time. Sign up for this, pump some 10fps video through it, let the charges add up, and write some "blockchain code stuff" (hand waving here) to process the same video using crypto-currency style resources. When you find something it missed or misclassified (whatever it takes to get the 10x refund) you file for your refund :) 

Though I would imagine they are only paying the refund based on the cost of processing that specific frame or set of frames that was missed, which might cap it at a buck or two. In which case, you could run some video through, watch the results of the system, find it's weaknesses, and then just pre-select video clips with valid objects likely to be missed to rack up refund charges.

 

To become a serious contender in video surveillance, they will need fundamentally reexamine how they approach the market.

Agreed thoroughly.

 
Agree
Disagree
Informative
Unhelpful
Funny

Hi there,

1. Fair use and fair play will be in effect, so 'gaming' it would break the agreement. The point is that we want to ensure we want to deliver value to the end user and this shows our commitment. If there's a valid scene that causes false positives or false negatives, we want to know about it and fix it for the customer.

2.  The cost stated here is for the Cloud API and cloud GPUs are simply expensive at the moment. There are ways to limit the number of frames needed to be analysed and we work with our customers on this.

On-premise pricing will be "per camera". No announcements yet as we are still working on several VMS integrations.

Agree
Disagree
Informative: 2
Unhelpful
Funny

Thanks for the feedback.

How do you envision this in terms of broader application? Even with using motion-detection as a coarse first level filter, it seems like this is a very resource intensive architecture, sending gobs of video up to the cloud to see if maybe there is a person in it? And then if there is, send something back down to the on-premise VMS? Plus, competing with all the other analytics offerings out there.

 

Agree
Disagree
Informative
Unhelpful
Funny

The cloud API is for systems already in the cloud.

For on-premise VMS, we have a local copy running and you'll just pay for a license, so you're not limited by the per frame cost. We are working various VMS integrations at the moment.

We'll be at ISC West to demo this, if you're there, do drop by. Everything we demo at ISC West will be on-premise only, since we're expecting poor Internet connectivity.

Agree
Disagree
Informative: 1
Unhelpful
Funny

Even at 1fps, it would be $3.60 per video hour, which over a month would still be $1,000 per camera.

Hey Sean - I think your calculations are incorrect. At 1fps (which is more than enough for cloud applications):

($0.10 / 1,000) * 3,600 (seconds in an hour) = $0.36 per hour.

Also to re-emphasize, this is Cloud API pricing. On-premise will be per camera.

Agree: 1
Disagree
Informative: 2
Unhelpful
Funny

Jimmy,

You are correct, we hadn't factored in the $0.10 for 1000 frames. I updated the report to reflect this.

Agree
Disagree
Informative
Unhelpful
Funny

Would it be possible to pipeline VCAs from another system into Image Intelligence?

IE: Run a false alarm clip from 'insert manufacture here' into the Image Intelligence to potentially weed out nuisance alarms? 

Agree
Disagree
Informative
Unhelpful
Funny

If you already know the clip is a false alarm, what benefit would you get from running it through this (or any other) VCA system?

Agree
Disagree
Informative
Unhelpful
Funny

Yes, this would be a good use case.

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

This article is part of IPVM's 6,956 reports, 927 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

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