Bosch IVA/EVA Video Analytics Tested

By Rob Kilpatrick, Published Sep 16, 2021, 09:05am EDT

Bosch has released the latest versions of their IVA/EVA analytics, with an overhauled calibration process and claimed accuracy improvements. But will these moves change their positioning versus Avigilon, Axis, Hanwha, or Hikvision camera analytics?

IPVM Image

To see how this new generation really performs and what has changed, we tested IVA on Bosch's starlight 8000i and MIC 7100i and EVA on the Flexidome 3000i, examining:

  • How do they handle common indoor and outdoor false alert sources such as foliage, shadows, animals, rain, and lights?
  • Are people accurately detected and classified when walking, running, crawling, or partially obscured?
  • What PPF is required for accurate detection?
  • How well does IVA PTZ auto-tracking work?
  • How has IVA/EVA calibration changed or improved?
  • How do Bosch analytics compare to other camera analytics like Avigilon, Axis, or Hanwha?

*******

** *** *******, *** and *** ********** ******** human ******** *******, ********, and **********, **** ** false ****** ** **** sources, ********* *******, *******, lights, ** *******, ****** both ********* ******* ** droplets ****** ** ***** rain.

*** ********** ********** ****** and ******** (********* ****, trucks, *** *****), *** EVA ************ ********** ******** as ******.

**** ********* ********** **** Exacq, *******, *********, *** Network *****, **** ******** searchable ** ******* *** Milestone ***** ***** ******** metadata ******.

*** *** *****'* ***-***** auto-tracking ********** *** ******** tracked ****** *** ******** in *** ***** ** up ** ~***' *****, following ** ****** ** interest **** ** ******** other ******* **** ** or ********* ******* **, though ******** *** **** if ******* ***** *********** behind ******* ** ** an ****** ******** ********** for * *** ******* before ****** ***** (~* seconds). *******, ***** **** caused ******** ** ****** objects **** ** ***********, wheels, ** *** **********, though ***** ****** **** not ******.

IVA ** *** *********** *** ********

**** *******, *** *********** (***** ** *****'* *****-**** models) *** ******** ****************** *** **** *****, ******** ***** ****** and ******* ******* ** false ****** ***** ********** triggered **, ********* ******* and ******/*******, **** ***** classification ****** *******, *** not **********. *** *** not *********** *** ******* in * **** ** testing, ***** *** ************* vehicles ** ****** ** least **** *** ***.

Versus ********, ****, ******, ********* *********

*****'* ****** ********** ** IVA **** *** ************* change ***** *********** ****** top ****** ********* **** as********'* ***,**** ****-******** ****** *********, ********'* ******* * **, ***** ***** ***** rejection ** ******* ** all ***** *********. *******, others ** *** ******* the *********** ******* ****** in *** (****** ******** and **** **** ******* optional ************), ******** ************ complexity, ****** *****'* *** wizard-based ***** ** ****** and **** *****-*****.

***'* *********** *** ******** moderately, ** ** ** longer ******* **** **** false ***** ****** **** in **** *****, ********** key ** ***** *** cost***** ****-********* ******, ** ** *** provides ****** *********** ************'* ******** ********* ********.

New *********** *******

*******, ***** *** ********** a *** *********** ******* that ***** ***** ******* setup, **** ******** *** each *********, *** ******** if *********** ******** ****** inaccurate, ****** ***** **** consistent *** **** *****-***** than ******** *****.

*******, ***** *** ******* is ******, **** ***** top-performing ********* (*.*. ********, Axis ****** *********, *** Hanwha ******* * **) do *** ******* *** manual ***********, *********** *****.

People ********** *******, ********, **********

**** *** *** *** accurately ******** *** ********** human ******** **** *******, running, *** ********, ***** below.

IPVM Image

**** ******** **** **** nearly **% ******** **** detected *** **********, **** as **** ****** ********* blocked ** * ***.

IPVM Image

False ****** ** *********

*** **** ********** ****** of ***** ****** ** IVA *** *** *** raindrops ********** ** *** camera's ****/** ***** ** the **** ****** ***** rain. ***** ***** **** detected *** ********** ** both ****** *** ******** in **** *********, ****** more ******** ** *** than ***.

IPVM Image

No ***** ****** ** *******

** *** *******, ******* IVA *** *** ********* false ****** ** ******* foliage ** *** *******, shown *****:

IPVM Image

No ***** ****** ** *******

************, ******* ***** *** small ******* (**** ** the ***, *****) ********* false ****** ** ****** IVA ** ***. *** animal *** ******** *** ignored.

IPVM Image

No ***** ****** ** ***** *******

******* ****** **/*** ** rooms *** *** ******* false ****** ** ****** IVA ** ***. ****** were ** ***** ** our **** ***** ** EVA.

IPVM Image

No ***** ****** ** *******

************, ***** ******* *** not ******* ***** ****** in ****** ********. ** with *******, *** ****** was ******** *** * bounding *** *******, *** did *** ******* ** alert.

IPVM Image

Person ********* ** ~*-* ***

****** ********* *** ******** at **** ~*-* ***, resulting ** *** ******** and ***** ** **** shown ***** *** * 1080p ********* *********. **** is *** ******** ** most *******, ***** **+ PPF, *** ******** ***** than **** ***/*** *****, which ******** *-* ***.

IPVM Image

IVA *** ************

*****'* *** ************* ***-***** ****** *** vehicle ************, ******** ***** to *** ** ******** rules ***** ** ****** positions, ***** ******* ************ upon **********.

** *** *******, ******** worked ****, ******* ********** when * ****** *** seen ** ** ***** zone, *** ******** ******** as ** ****** ********** the ***** *** ** a ******** ** ****** 500', *****.

****** ******** *** ******** even **** **** ** the *******'* **** *** obscured, **** ** *** person ******* ****** *** car *****:

******** **** **** ********** tracked ** *** ******* lot *******, ** ** ~15-20 ***. ******** *** be *** ** *** specific ****** ********** ** the ******, *.*., **** people, **** *****, ***.

No ********* ******* ******* ******* *** *****.

*********** ******** **** ****** the ***** ****** ***** triggered *** ****. ** does *** ***** ******** objects/switch ******* **** ** multiple ***** *** ***** of ****, ********* "******" to *** ******** ********** object ***** ** ****** the ***** ** ** stationary *** * ****** of ****.

Stationary ******** *** *******

**** * ******* *** stationary *** *** ****, tracking ****** ******* ** started ******** ******* ****** in *** *****. **** could ** ********* ***** by ********** *** ******** idle **** ** *** PTZ ***** ** *** area *** ****** ** idle ***. ***** *** person ****** * **** stops *** ******* ****** begins ****** *** ** tracked.

Swaying ******* ********* *******

******* ******* ** *** scene **** ** * flag ********* ****** *** PTZ ** ***** ** for * **** ******* of ****.

IPVM Image

***** ******** ** **** the **** *** **** a **** ******** ***** areas ** *** ***** to ***** ********** ******* and ******* ***** ******.

Some ******** ****** ****** ***** ****

*** ******** *** ********* accurate ****** ***** ****, which *** ********* ******* numerous ***** ****** ** other ************ ****.

*** *******, *** ******* and ******* *** ********** tracked ** *** ***** below, ******* ******** ** front ** *** ******'* lens.

*******, **** **** *** extremely *****, ******** *** triggered ** *********** ** cars, ******, *** **********. In *** **** *****, after ********** ******** * subject ** * *******, the ****** ***** ** and ****** ** ***** a ********** *****'* ********* as *** **** ******* to ***** ******** ** front ** **.

Solid ******** *** ***********

******** ****** ********** **** Exacq, *******, ******* *****, and *********, ***** ******* and ******* ***** **** integrated ****** *************** ** their ****** (*******, ******, bike, ***.)

IPVM Image

************, ******* *** ********* offer ******** ****** ***********, which ****** ***** ** search *** ******** *******/***** in ******** *****, ******* in ****** ******* / ********* ******** Search ******.

**** ***** (**** **** report) ********* ******** ****** in *********:

Versions ****

*** ********* ******** **** used ****** *******.

  • *** ** ********* *****: 7.81.0060
  • ********* ** ********* *****: 7.81.0060
  • ********* ***** *****: *.**.****

Comments (3)

**** *********** *******, ***** you.

Agree: 5
Disagree
Informative
Unhelpful
Funny

***** ******.

Agree: 3
Disagree
Informative
Unhelpful
Funny

****** *** *** ***** and ****** *******; * did **** **** ***** "test-driving" **** **** *** had **** ******* ********. I'd *** **** ****** Portal ************ ** **** and ** ***** ******** intuitive. ** * **** more **** *** ****** Portal ******, *'** **** back **** *** **** a ******* ** ***.

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

This article is part of IPVM's 7,270 reports and 968 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