Camera Analytics Shootout 2020 - Avigilon, Axis, Bosch, Dahua, Hanwha, Hikvision, Uniview, Vivotek

By: IPVM Team, Published on Jan 22, 2020

Analytics are hot again, thanks to a slew of AI-powered cameras, but whose analytics really work? And how do these new smart cameras compare to top non-AI analytics?

In this report, we compare and contrast 15 different camera analytics from 8 manufacturers from our tests:

Inside, we cover:

  • The best, average and worst performers overall
  • The top 3 best performers compared in detail
  • The weaknesses of the average and worst performers
  • How the competitors resisted outdoor false alerts (e.g., foliage, small animals, shadows/lights, rain)
  • How the competitors handled indoor false alerts (i.e., shadows, lights, and flashes)
  • How well analytics detected subjects running and crawling
  • How accurately objects were classified as vehicle, human, etc.
  • Who delivered the greatest range (ranking pixel density / PPF requirements)
  • What specific detection issues each analytic had
  • Who was better or worse at configuration and calibration
  • How VMSes supported (or not) these cameras analytic events

********* *** *** *****, ****** ** * slew ** **-******* *******, but ***** ********* ****** work? *** *** ** these *** ***** ******* compare ** *** ***-** analytics?

** **** ******, ** compare *** ******** ** different ****** ********* **** 8 ************* **** *** tests:

******, ** *****:

  • *** ****, ******* *** worst ********** *******
  • *** *** * **** performers ******** ** ******
  • *** ********** ** *** average *** ***** **********
  • *** *** *********** ******** outdoor ***** ****** (*.*., foliage, ***** *******, *******/******, rain)
  • *** *** *********** ******* indoor ***** ****** (*.*., shadows, ******, *** *******)
  • *** **** ********* ******** subjects ******* *** ********
  • *** ********** ******* **** classified ** *******, *****, etc.
  • *** ********* *** ******** range (******* ***** ******* / *** ************)
  • **** ******** ********* ****** each ******** ***
  • *** *** ****** ** worse ** ************* *** calibration
  • *** ***** ********* (** not) ***** ******* ******** events

[***************]

Performance ********

***** ** *** *****, analytic *********** ***** **** four ******:

  • *** **********:********'* ** ****-******** *********, Bosch ***, *** *****'* Analytics+ ******* ********* **** in *** *******, ************ rejecting ***** ****** **** common ******* ******* *** out, **** **** *****/***** than ******* *** ************.
  • ***** *******:********'* ** *********, **** Perimeter ********, ********* ********** and ********, *** ******* SMD *** ********* ***** average ** *** *******, rejecting **** ***** ******, though **** **** *** limitations. ********'* ** ********* drastically ******** ********* ******** at *****, *** ** issue ** ** *******. Axis ********* ******** ************ suffered **** ***** ****** caused ** ********** *** occasionally ********* ** ****** objects **** ** ****** doors *******/*******. ********* ********** and ******** **** ******** from ***** ****** ** small *** ***** *******. Finally, ******* *** ********* well, *** **** ******** people, **** ** ****** for ********, **** ****** PPF ************ **** ******.
  • ******* ***********:****' ****** ***** *** VMD4, *****'* ********* ***** Analytics (***), *** ******'* Wisenet * ********* *** performed *********, **** ******** false ***** ******, *** a **** ***** ********* analytics **** *****, *********, and *******.
  • ***** ***********:*****, *********, *** ******* analytics ************ ********* ***** in *** **********, **** increased ***** ***** *****, shorter ********* ******, *** fewer *** ************. ***** analytics *** ****** ** be **** ** *** to *** ******* **** these *************' ****, ** they **** ********** ******* "dumb" ****** ***, *** for ******* *** *** integration **** ***** *******, they *** *** ********* options.

Top ********* ***************

******* ** ********'* **, Bosch ***, *** ***** Analytics+ ** **** ******, there *** ******* *** differences ******* *** *** performers ***** ****** *********, detailed *****.

  • ***********, ********* ******** *** Dahua:********'* ********* *** *********** "self ********", ********* ***** simply ** **** ****** a **** *** **** a ********* **** ** many/most *********. *********+ **** requires ***** ** ****** the **** *** ****** people ***/** ******* *********. By ********, ***** *** requires **** ******** *****, with ***** ******** ** measure *** ****** ** objects ** *** *****, and ** ****, ******* objects, ******** ********* *** angles ** *** ****** plane, *** ********, ***** length, *** ******** ****** entry (********* ** *** video *****).

  • *** ***********, ********* *****:***** *** ********** **** multiple *****, ********* **** events *** ******** *** metadata. ********'* ********* ********* only **** ******** ******* Center, **** ** ************* with *****, *******, *********, or ******. ***** *********+ with ********* *** ******* Optix' ** *******, *** not ********, *****, ** Genetec.
  • ****, ********* *****:***** ****** *********+ ******* at **** ***** ****** than ******** *** *****, with **** ******* **** models ******* *** ~$*** online, ******** ** $***-*** at * *******.
  • ********* *********/*** ************ *******:** **** *****, ********'* H4 ******* ******* ********* distance ** *****, *** their ** ****** ** not ****** **** **** limitation. *** ***** *** analytics **** ****** ******* PPF ************, **** ~*-** PPF.

******* ** ***** ******, these ********* ******** ******* false ***** ***********, ********* most ****** ****** *** outdoor *******, **** ** blowing *****, ***** *******, rain, *******, ******, ***.

Deep ******** ********* ****** ** *******

** *** *******, *** deep ********/**-***** ********* ********* above ******* ** ****, dominating *** *** *** categories, **** ********'* *** and ***** *** *** only ********* ***** **** not **-***** ** ***** top ********. ** ****** that ** ** ******* availability ********* ** ******** and ********** ********, ** will *** **** ***** continue, **** **** *** cost/"cheap" ** ******** *********** superior ** **** ***-** analytics.

Outdoor ***** ***** *********** **********

**** ********* ****** *** issues **** ******** ***** alert *******, **** ** foliage/brush, ***** *******, *******, and ****, ******** *****, though ******** (**** **/**), Bosch ***, *** *****'* Analytics+ *** *** ****** from ****** ** *** of ***** *******.

******* *** ********** *** most ****** ***** ** outdoor ***** ******, *** in **** **** ********* were *** ********* ** it. *******, *********'* ********* Analytics *** ******* ********* all ******** **** **** constant ****** **** ******* was ******.

**** ** *** *** most ****** ****** ** false ****** ** ********* outdoors, ******* ***** ****** in **** (****** ***** and ****), ***** ***, and *****, ******, *********, and ******* ********* *********. Note **** ***** ****** were ****** ****** ** droplets ******** **** *** cameras' *****, *** **** alerted ** ***** ******** itself.

****, ***** *** ***** animals ************ ********* ****** in **** ****, ***** EVA, ***** *** ******* intrusion, *** *** ********* analytics, ********* ********** *** AcuSense.

**** ****** ***** *** triggered ** ********** ******* on ******* ** *** scene (**** ** ****** car *******), *.*., *** trailer ***** ** *** example *****.

** ******** ** ***** common *******, *******'* ********* analytics **** *** **** ones ** ***** ** shadows ****** ** ****** passing ********, ***** *****. No ***** ****** *** these ******.

Indoor ***** ***** ***********

** **** *******, **** analytics ** *** ***** alarm ** ****** ****** sources, **** ** ****** being ****** ** *** off, *********** *** ********, and ******* **** ** other *******, **** **** notable **********.

******'* ******* * ********* were ***** **** ***** issues, ******** **** ****** were ****** ** ** off, **** ** **** scene:

*** ******** **** ****** were ****** **/***, ******* light **** *** ****, below:

Detection *** ************** ******

****, *** *** *** analytics **** **** ** reliably ****** ******** ******* and ********, ******** *****. Additionally, ************** ****** (** those ***** ********** ******* as ******* **. *****) were ****, *** ******* in ****. *******, **** analytics ******** ****** ****, resulting ** ******** ******.

*** **** ******** ***** missed ******* ******** ** our ***** *** *********'* AcuSense. **** ******** *** and *****, ******* *** out, ********** ** ******* the ******* *** ****** the ***** ** **** or ******.

**** ********* ******** *** unable ** ****** ******** subjects ****** * ******** setting *** *******.

*******, ***** **** ******* allowed ******** ******** ** be ********, ** **** resulted ** ***** ****** on ******* ** *** scene, ***** ****:

[optimize output image]

** ********* ********, ****** walking ******* *** *****, mostly ** *** ********** and ******* ****** *** camera **** ********** ********** as ********. **** ***** lead ** ********** ****** detections ** ***** ****** to ****** **** ******.

*******, ** ******* ***** AI ***** ********* (******** H5A, ***** *********+, *** Hikvision ********), ****** ******** in *** ***** ********** triggered ******. ***** ****** were *** ******* ** any ***-** *********, ** the ******* **** ****** not **** ** ******.

Constant Detection Of Parked Cars

PPF ************/********* *****

*** ****** *** ******** by ********* ****** *** ~5 *** ** **** Perimeter ********, *** *** night. *******, ********* ******** was ***** ** ***** alerts **** ******* ******* such ** ***** *** small *******, ** ****, unlike ******** *** ***** IVA, ***** ******** ~* PPF (****** *** ******** in **** ******, ********* below) *** *** *** suffer **** ***** ******.

*****'* ***/*** **** **** lowest, ** */* ***, followed ** ********'* ***, Axis ****** *****, ********* AcuSense *** **********, *** Vivotek ***, *** ********* 10-12 ***.

********'* ** ********* *********** a *********** **** ** detection ***** ***** * lux, *** ***** ** other *********. *** ******** jumps **** ~* ****** the *** ** *** 5 *** * ** models ** **+, ******* detection ******** ***************. *** 3MP ****** ********* ********** better, *** ***** *********** nearly * **% ******** in ********.

Avigilon **

******* *** ******* ***** clearly *******, **** ************ human ***** *** *******, Avigilon ******* **** *** able ** ****** *** until ** *** **** nearer *** ****** **** other ******. *** *******, at ~**' *****, *** subject ** ******* *****, but ** ***** ** displayed:

** ********, ********'* *** cameras ******** ** ~*-** PPF **** *** *** night:

9-10 PPF Required For Detection Day And Night

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

*** *********** ****** ******, with **** ********** ******** events, ****** ********** ******** boxes, *** ****** *** supporting *** *********** ** all, ********** ** *** chart *****:

  • ******** ******* ******:********'* ****-******** ********* **** supported **** ** ***** own ******* ****** ***, with ** *********** ** third *******.
  • ***** ***/**** ********* ********:***** *** *** **** Perimeter ******** **** *** most ********** ******** ********* tested, **** ******** ***** supported ** ******** *****
  • **** ***** *****/**** *** Hanwha ****** *********:****' ***** ********* *** Hanwha's ******* * ********** events ** *** ***** tested ****** *** ******** Control ****** (*** ** bounding *****).

Axis **/***** *** *** ***********

****' ********* ******** *** Bosch *** **** *** only ****** ********* ***** integrated ******** ***** ** well ** ****** ** third ***** *****, ****** the ***** *********** ****** slightly. ********* ******** ******** not **** ******** *****, but *****, ** **** as * *****/*** ***** indicator ** *** *** right ** *** *** window, ***** **** ** Milestone:

** ********, ***** *** displays **** ******** ***** around ******** *******, **** no **** ***********, ***** here ** ******* ******** Center.

****/************

**** *** ** *** tested ********* **** ******** license/channel *****: **** ***** Suite (~$**/******* *** ***-* series **** ******) *** Axis ********* ******** (~$***/*******). Others *** ******** ** no ********** **** ** a ***** *************' *******.

*******, ****** ***** *** be "** ********** ****", some ************* **** ******* analytics ** ****** ****** models, **** ** ********'* H5/H4 ****** (*** ******* mini **** *** ****) or ******'* ******* * instead * (***** ******** VMD ****). ** ********, Dahua, *********, *** ******* include ********* ********* ** nearly *** ******, **** few **********, **** *** problem ***** **** ***********.

Comments (17)

*** **** '**** ********* Defender' ****** *********** ** the ******** ***** (*******) versus *** ******* **** (Above *******). ****, **** you ** ****** * list ** ****** ******?

****** *** ******** **** out, * ***** ********* defender ** ** ********* show *** ******* ** the *****. ** *** would **** ** *** the ****** ****** **** are ** *** ********** reports ****** ** *** top.

** ***, ****** ****** attention * *** *** that **** ** * summary ** *** ******** tests, ** ******* ** including *** *******. **** do *** **** ***** the *** ********** ********* models, ** ********? * see **** **** ********* Global *** ****** **** several **** ********* ****** than **** ** ********* in *** *.*., ********* 4K ******** **** ** this ***:

**-*********-***

***** **** *** "**" in *** ***** **** (which *** **** ******* since ***** *** ******* "G0" ******), * *** no ********** ** ******* it ** *** ***. Hikvision ** ******** ****** false ***** ********* *** the *** ***, *** in **** **** **** have * ******** ******* note *** *** ******** Accusense ****** ******** **** it ******* ***** ******.

**'** ***** **** **** >2000 ****** ** ** for ************ **** ******** and *** ********* **** happy **** *** ******* of *** ***. *** that ** ** *** available ******* ** ******** to ** *** ******** both ******* (** ********* megapixel ********) ***** **** to **** *****. *** unexpected ******* ** *** test ***, **** *** H5 *** ******* ************* worse **** *** ** during *** ***** (**** external ** ******) ** detecting ****** ** ********* clothings *******, ******* ** crawling ******* *** ******* in ********* ********* (**,**,**,**.** meter).

**** ********** *** ******* with *** ******** ** got *** ********* ********* back:

"**** *** *** *********** and ***'* ***** **** the **** ********* *******, which ** *** ***** perception ** *** ** cameras'confidence ****. For a good reason, this value is critical to find out where the technical boundaries of our system lies. After evaluation of the clips, we quickly became aware that we missed to properly communicate the right background information on how to actually read this value correctly. In first glance, it looks as H5 confidence rates have no improvement to H4, but H5 is actually using a completely different scale. Simply put, a low H5 confidence rate equals to a mediocre H4 value. To put it in clear figures, a 50% H4 confidence rate equals to 10% in H5.

********* *** ******* ***** with **** ***********, ** can *** **** ** gives ** ************* ****** object ********* ******* *** thus, ******* *********, *** less ***** ******."

****** ******* *** **** extensive **** ******, *'* like ** ********** ** someone ** *** ***** the **** *****. *********** I'd **** ** ********** what ** *** ***** of *** ********* ** our ********.

****** * ***.

** ***** ******, ** looks ** ** ********** rates **** ** *********** to **, *** ** is ******** ***** * completely ********* *****. ****** put, * *** ** confidence **** ****** ** a ******** ** *****. To *** ** ** clear *******, * **% H4 ********** **** ****** to **% ** **.

****** **** **** ** me. **** ** *** "different *****" *******?

**** * ** ****** says **%, ****'* *** reality? ** ** ***** half *** **** ** one *** ** ***?

** **** **** ** me. **** ** *** "different *****" *******? **** a ** ****** *

**** **** **** **% of ********** ***** ** H4 ** *** **** level ** ********** ** as **% ***** ** H5.

**'** ********* ******** *** clarity ** ****. *'** post **** **** **** they *** ** ***********.

*** **** ******* *** pretty ********* **** *****. H5 *** *** ****** H4 *** ********* *********. In *** ** *****, subjects ***** **** **** and *****, ******* *******, and *** ** ****** up ** ***.

*** *** *** * feedback **** ******** ** the *****, ***?

******** *** *******, *** confidence ******* ** *** H4 *** ** ***** directly ********** *** *** only ******** *** ** compare *** *** ** to ******* ***** ********* and ****** **********:

***** "**********" ************ ***** the **** ***** (*** certain ** *** ********** that * ************** ** correct?) *** **** ** and **, **** *** not ******** **********. ** leverages * **** ******** convolutional ****** ******* ***** classifier **** *** ******* with * ****** **** set **** *** ** classifier. ********** ****** *** calibrated **** **** ** the ****** **********, ** maximize ********* ***** ***** keeping ***** ********* ** an ********** **** ***** at *** ******* **********, we ******** *** ***** positive ***** ***** ******* detections ** ** ********** rate. * **** ******** classifier (**) ****** **** correct ********** ****** *** confidence ****** ******** ** a ****** ********** (**). Thus, ******* ** *** difference ** *** ********* range ** *** **********, it ** ********* ** directly ******* ** ** classification ********** **** **** of *** **.

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

** ***** ********** **** from ******** ** ****, when **** ***** ** explain *** **** ******** of *** ** ******** to *** **. **** have **** **** ** the ****-***** *** ******'* explain *** *******, ***.

*******, ** * ********** your ***** *****, ** can ******* *** ** and ** ** ******* the ********** ****** ** a ******* **** (***: 30%) *** ***** *** missed ********** *** ****.

** ******* **** ****, the ** ********* ** far ****** **** *** H5. *** **** ** actually ** *******, ******* H4 ** ** ****** available *** ** ****** an ***** *** ** people.

*** ******* **** ** can ** ?

*** ***** **** *** thought ***** *** ******* the ******* ********* ********* with ******* *******?

*** ********, *** *** testing **** **** *****/**** targets ****** **** ** their ***** *** *****?

****** *** **** **********, we *** ******* **** Uniview *** ******** ****** their ********* ******** **** much (****) ****** *******, might ** ****** ****** for ** ** *** that ***** ** **********...

*** **** **********, ** are ******* **** ******* and ******** ****** ***** Alphaview ******** **** **** (much) ****** *******, ***** be ****** ****** *** us ** *** **** added ** *

***** ***** **** *** you ****, *** ?

** ***** - ** the *** ************ **** the ****** (******* ********) of *** + ** blue ** *** ***** but ****** ** *** key - ** ****'* my **********.

***** ****** *** ******* to *** *** !!!

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

** *** ** **** some ****** *** ******* coverage, *** *** **** false ****** *** ****** unreliable. **** ******* ****’* Mitch **** *** ** just ***** *** ****** to ******** **** ********** results.

*****,

**** *** **** *** kind ** **** *** "group ** ******"?

*** **** * ***** of *-* ******, **** are ******* ********, ** being ******** ** * vehicle ** * ******?

**** *********** ******* ** the ***** ** *********. I **** ***** *** "Performance" ****** *** *** "Smart" ****** (**** **** release?), *** *** **** video ********. (** ******* subscription ****.) **** *** the $*** ****** **** but ******* *** **** than $*** **** ******** camera ** ******* **** the "*****" ****** **** on **** *** ***** detection. *** ***** *********** series **** *** ** well *** **** ****** with * "**** *********" option, ***** ** **** not **** ** *** even ** ******* *********** and **** * **** 5 ** * **** in ***** ** *** camera. ********* *** * MP ***** *** **** miss ***** ************, *** then *** * ** smart *** **** ************ trigger ** * ***** with ** ****** ** all **** ** **.

* ***'* ****** *** teardowns ** *** **** so * ***'* **** what ***** *** ****, but ** ***** ** interesting ** **** ** in ***** ***** ********* against *** **** ****** dollar *******. **'* *********** moving ****** ** **** or **** **** *** camera, *** * **** yet ** ** **** to **** ** ****** with * ***** ****** boxing *****. ******** ****** you ***** **** ** get * ****** ** it ****** ** ** your **** *******.

Login to read this IPVM report.
Why do I need to log in?
IPVM conducts unique testing and research funded by member's payments enabling us to offer the most independent, accurate and in-depth information.

Most Recent Industry Reports

Masks Cause Major Facial Recognition Problems on Feb 24, 2020
Coronavirus is spurring an increase in the use of medical masks, which new IPVM test results show cause major problems for facial recognition...
Every VMS Will Become a VSaaS on Feb 21, 2020
VMS is ending. Soon every VMS will be a VSaaS. Competitive dynamics will be redrawn. What does this mean? VMS Historically...
Video Surveillance 101 Course - Last Chance on Feb 20, 2020
This is the last chance to join IPVM's first Video Surveillance 101 course, designed to help those new to the industry to quickly understand the...
Vulnerability Directory For Access Credentials on Feb 20, 2020
Knowing which access credentials are insecure can be difficult to see, especially because most look and feel the same. Even insecure 125 kHz...
AI/Smart Camera Tutorial on Feb 20, 2020
Cameras with video analytics, sometimes called 'Smart' camera or 'AI' cameras, etc. are one of the most promising growth areas of video...
China Manufacturer Suffers Coronavirus Scare on Feb 20, 2020
Uniview suffered a significant health scare last week after one of its employees reported a fever and initially tested positive for coronavirus....
Cheap Camera Problems at Night on Feb 19, 2020
Cheap cameras generally have problems at night, despite the common perception that integrated IR makes cameras mostly the same, according to new...
Milestone Launches Multiple Cloud Solutions on Feb 18, 2020
Milestone is going to the cloud, becoming one of the last prominent VMSes to do so. Milestone is clearly late but how competitive do these new...
Video Surveillance Architecture 101 on Feb 18, 2020
Video surveillance can be designed and deployed in a number of ways. This 101 examines the most common options and architectures used in...
UK Stands Behind Hikvision But Controversy Continues on Feb 18, 2020
Hikvision is exhibiting at a UK government conference for law enforcement, provoking controversy from the press, politicians, and activists due to...