You mean I can’t hit the “CSI video enhance” button to get the LPR software working correctly to get that car 200ft away travelling 50 miles an hour from the reflection of this tea spoon? Geez we can put a man on the moon...
Video Analytics Usage Statistics 2018
Published Mar 09, 2018 14:57 PM
Video analytics is hot again (one of the top 2 trends going into this year), this time driven by deep learning and artificial intelligence.
But, how much of video analytics are actually being used today?
200+ integrators told IPVM how often they use it and why they either choose to use or avoid analytics. And there is a new #1 reason why integrators say they do not use analytics. In this note, we break down the results and compare to our 2011, 2014 and 2016 studies.
2018 Usage
Almost half of integrators still rarely or never using video analytics with less than 1 in 5 using them often:
******* *********, ********** ******** ***** ** effectively **** ******** ** ****, **** just * ****** ******** ** *** overall ****** *** ******:
#1 - **** *******
** ****, ***** *** * *** number ***, ***** *** *** * material ****** ** **** ********** *******. The **** ****** ********* *** **** costs *** ***** *********, ** **** sampling ** ******** ********** ********* *********:
- "*** ******** ** ****** * ***** customer."
- "*** ******* *** ** *** ****."
- "****** *****'* *** *** ** *** they ******* ***'* **** ** *** the ****** ***** *** **. **** have *** ******* **** ** ********* happen **** **** ** **** *** just **** ** **********."
- "*****. **** *********** *** ********** ** lowest ***** ******* */* **** *** only ******* ** *****."
- "**** ***** ************* ****"
- "********* ** ****** ********* ** **** secure ********. *** ***** ******** **** us ******* ******."
- "****** *********** *** ****"
- "** ***** ** ****** ******* ** the ********* ****** ** ******* *** having ** *** *** ****."
- "*** **** ** ********* *** ******* requires ** **** ***** **** ******** recording."
- "**** **. ******** *********** ** ****."
- "**'* ***** ********* *** **** *** get"
- "**** ********* ** *** *** *** software ***** ***** *** *****"
- "** **** ******* ********* *** ********* have *** ******** *** ***** *****"
- "* **** **** **** **** **** to **** *** ********* *** **** from *** ********* **** ** **** really **** **** ** ***** ** helpful ** ***** ***** **********."
- "*** ***** ****** ********, ** ********* do *** **** **** ********* ** keep *** **** ****."
- "** *** **** **’** ***** ** too ****** ** *** ********* ******* and *** ****** **** ** **** to ** ********* ***. ******** **** more ********* (*****, ***********/****, *******) ***** edge *****, *** ****** **** **** back **."
Race ** *** ****** ****** ** ***** *********
**** ** *********** ***** **** ** that ******* ***** ******** ******* *** at ***** **** ****, ** *** somewhat ********* **** *** **** *** years. *** ************** ** **** ** that ** ** ** ******** ****** of *** **** ** *** ******. With ****** ****** ********* ** *******, it *** ************* **** ***** ********* look **** **** *********. ** *** integrator ********, "**** *** ***** ***** rather **** **** ****** ******** **** spend ******* ** *********." ** *** are ******* $*** *******, ****** ** ~$500 ***** ********* *** ******* **** more ********* *** *********** * ***** investment **** **** ****** **** ******* (at ***** *** ***** **** ******* camera ****** *** **** ******** *****).
Unrealistic ********** ********
*** ***** *********** ********* *** ***** customer *********** ************, ********* *********:
- "******** ************; **** **** *** ******* for ******* ******."
- "******** ************ *** *** *********** ** analytics ***** *** ******* *** ********* performance ** **** *********. ** ** can ******** ********* *** ********'* ************ for ***** ****** *** ******* ***********, then ** *** **** ******* ** deploy ******** ********* *** **** ********."
- "*********** ********** ** *** ****** **** is * *********."
- "** *** ******* ** *** ********'* expectations *** *****."
- "**** ****** ********* *** ****** *** than **** **** ****, * ***** don't **** ** **** ****** * customer *** ** ** ** *** performing *** **** ******** ** **** probably ***** **** **** *** *** analytic ******."
- "********* *********** *********** ********** ** *** you **** **."
- "******* ********* ** * ******** **** expects *** **** **** ** *** get *** **** * ******** *********. Even ******* **** **** *** ************ ahead ** **** *** ***** **** disappointed."
- "* ********* *** *** ******* *** to **** *** ********'* *********** ***** may ******* ***** ** *** *******."
- "***** *** "*** ******" ** **** had **** ********* **** ** ** with ***** ** ***** **** ***** the ****** *** ******** ********* **** can ** ********** ******** **** ***% accuracy *** ** ****** * **** deal ** ********* ** ********* *** customers ***** *** *******"
**** ** *********** ***** **** ** customers **** **** *** *** ****** when ** ***** ** ******* *** resolution *** **** ** ****** ** objection ** ******* ** ******* ******* as ***** ****** **** ** ***** integrators ********* ** *** **** ***** analytics, *** ** **** **** ********* cameras.
*** ************** ** ****, ****** ** cameras, ***** ********* *** **** **** some ******* ***********, *** ******* *********** of ***** ********* (*.*., ***** ********* and ****** ******) *** ***** *** high ** ******* ****** *** ***** for **** ** **** ************.
***** ****** / ******** ******** *******
** *** ****, ***** ****** *** accuracy ******** **** ********* *** **** commonly ***** **********. ** ****, **** declined, ****** ***** ***** ** ******** integrators. How **** ** **** ** *** to *********** ****** ********* *** ******* better ** *********** ********* **** ** pricing ** *** ***** *** ** is, ** *****, ******** ** * positive ****.
Demand ** *****, ****** ** ***
*** ******** **** *** ***** ********* is **** *** ****** ** *****. Integrators **** ** *** ***** *********. *******, the ******, ******* *** ****** ** offerings *********, ***** ******* *** ********** are ***** *** *** ******* *** levels **** *********** ****.
*** *** ******** **** ******* ******* and **** *** ***** ********** ** deep ******** ********* ****** ****** ****** (readily *********, ********** ******, **** ****** quality) ** ***** **** ******** ** video *********.
Vote / ****
Comments (14)
UI
Undisclosed Integrator #1
Mar 09, 2018
Sean Patton
Mar 09, 2018
Brian Rhodes
Mar 09, 2018Sean Patton
Mar 09, 2018
Campbell Chang
Mar 11, 2018
JC
John Collings
Mar 13, 2018UM
Undisclosed Manufacturer #2
Mar 09, 2018Very difficult to answer that question because "video analytics" means many different things to many different people.
I ask the question all the time to integrators and end users both, "what is your understanding of "video analytics?" The answers are wide and varied;
"that's bag left behind, right?"
"analytics are used to report license plates to law enforcement"
"it's just a virtual trip line"
"glorified motion detection"
"it can tell you age and gender and count people"
etc
A camera is a verification device (live or after-the-fact when recorded). Video analytics is fundamentally using software/hardware to allow a camera to act as a detection device. What you're "detecting" can be wide varied as noted above.
In order for "video analytics" to achieve mass adoption there must be specific use cases that solve a real world problems and provide real world, verifiable benefits. Analytics must have a very high probability of detection/accuracy and a very low nuisance/inaccuracy rate for the given use case or it will fail (as demonstrated in the past).
All the talk about deep learning, machine vision, artificial intelligence, etc is gibberish to the masses when it comes to real world business cases and justifications for investing in it.
The world first needs to understand where a given analytic engine can be used to either detect events to prevent crime, provide intelligence to improve operations/efficiency/sales or significantly reduce investigation time, et al.
When those business/use cases can be clearly defined and substantiated, there will be wide adoption for a given analytic function. Until then, not so much. Get the right information to the right people at the right time to succeed.
Mark Palka
Mar 12, 2018we run a station for 14 years that does video verification using analytics and although there are issues there are 2 things that should be mentioned 1- analytics is getting better and better and 2- it is much better than VM or motion sensor, and human beings who go brain dead after 45 minutes.
In our town there was a major fire recently and the site was secured with cameras and operators with VM as the trigger - the human operators missed it because too many false alarms. easily explained in this video:
Analytics is not perfect but in many applications it is the preferred choice by far.
JH
John Honovich
Mar 12, 2018Mark, thanks! What analytics are you using? Just curious to see what you have seen directly getting better.
Mark Palka
Mar 12, 2018Hi John we have used Object Video and Activeye maybe 10 years ago and Aimetis more recently but we have been frustrated with constant upgrading fees etc. So we have used our own open cv which we don't charge for. All we use it for is detection, the accuracy is much better than it was but we have also understood it more. For example it works much better where we have a contrast in the grey scale. For example white on white or black on black is no good. We have found it to be very accurate using thermal.
In a station, monitoring a people access car lot using VM and lets say 12 cameras you will process 1000 alarms a hour minimal on any given night this puts an operator to sleep. By contrast for the same size system using analytics and proper site conditions and camera palcements it will be 8 to 12 a night. Making it manageable for a business case. In the past year we have maybe missed 3 or 4 alarm conditions on approx 1200 cameras. In some cases this was a known condition and a result of customer budget and in others the event was caught on another camera.
the bottom line is nothing is perfect but this is the best for a open lot or outdoor lot concept
UM
Undisclosed Manufacturer #3
Mar 12, 2018it is much better than VM or motion sensor,
By motion sensors I guess you mean things like PIR, right? I've always wondered how video analytics does compared to PIR or microwave.
Mark Palka
Mar 12, 2018yes PIR the issue i believe and I guess alarm guys can correct me if I am wrong - but PIR's need to be terminated and not by the ground shooting them into free air makes them unstable. Not sure about microwave but i have heard a lot about wi fi as being a next way to monitor outdoor sites
Rafael Oneda
Mar 10, 2018I'm not a enthusiastic of IVA... actually I think that it is a great way to do a frustrated end user!
The most embarrassing situation is when a concurrent integrator or manufacturer makes the customer's expectations high, claiming that they can get a high accurancy in any conditions, using a low-price camera, etc... I need to disagree with this and frustrate the customer, so, some customers think about me: "your solutions are bad, becouse cant do the amazing things showed to me by other company and, besides that, is more expensive... get out of here!!"
Sure, after deployment, he probably will be frustrated, but, I've already lost the deal.
Is almost a ethical issue.
UI
Undisclosed Integrator #4
Mar 12, 2018They all seem to work great in sample videos and PowerPoint. In the field, not so much.