Hikvision Dual Lens Face Recognition Camera Tested

By Rob Kilpatrick, Published Nov 19, 2019, 09:35am EST (Research)

*********'***** **** ****** *****************, ****** **** ** "****** ******** deep ******** ********* *** ******** *** to ******* ******* **** *********, *******, and ********** ** ********** **** *********** and ******* ***** **********."

** ****** *** ****** ******-**********-*, *********:

  • **** ** **** ****** ******* ******* the *****?
  • **** ** *********** ******?
  • **** ********** **** ** ********* *********?
  • **** ** ********* ********* ******* ** faces?
  • **** *** ** ******** *** ******* performance?

****: **** ****** *** **** **** in ***** *** ***** ***** ** the ***** *** * **** ** more *** *** **** ******** ******** for **** ** *** ****** ******.

*******

*** ********* **** **** ******'* ****** recognition *** ****** ******** ***** *********-*********** settings, **** ** ********* ************ (*** person ********** ** *******) ** **** a ***** ** *******, *** ****** performance ** *** ****** ***** **** been ****** *** ***** **** *** systems.

*******, ** ********** ******* ********* ******* as ***** (*** **** *** *********), suffers **** ****** ***** *** ******* confidence ** **** ******, *** ***** range ******** ** ***** ******* ****** (~10-11') *** ** *** **** ***** lenses **** ** *** ** ********* view ** *** ******** ******.

Key ********

** *** *******, *** **** **** Facial *********** ****** *** **** *********:

  • ** ***** ************:********** * ***** ** *******, *** camera ***** ************* *** ****** *** another ****** ********** *** ****** ****** and ~*** ***** ** *** ****** libraries.
  • **** ********** *** ***********:** * *****-****** ***** **** *******, the ********* **** **** ****** ********** subjects **** *** **** ********** ** evenly *** ******. **** ***** *** historically **** ** ***** *** ***** face *********** *******/*******, **** ** *****, AnyVision, *** *******.

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

  • ********* ******* ******** ** *****:******* ** *** ***** **** ** boxes ** **** ***** ******** ** floor **** ******** ** *****, ******* iVMS's "**** ******* *********" **** **** useless **********. ***** ** ** *** to ****** ******** ******* ** ****** them **** ********* ******* ******** **** area ** *** ***** **** *** detection ****.
  • ***** ****** ** ****/** ******:** * ****/** ***** (~*.** ***), ~1/6 ** ***** **** ****** **********, even **** * *** ******** ****** of ~*.*** *** * ******** ** ~10°.
  • **** ******** ******* *******/*********** ***** ** IR/Light *******:***** *** **** ******** *******, **** in ** ** ***** *** ***** may *** *** **** ** ****** faces ******* *******, **** ****** ****** this ******* ****** *** ****** ***********.
  • ***** **** ********** ******** ***** *********** Settings:***** *** *********** ** ****** (**** pupil ** *****) ********* ******** *** only ~**-**' (~** - *** ***). Pixel ******* ****** *** ** ******* to * ******* ** **, *** detection ****** **-**' ** **********.
  • ** ******** *********:******* * ******* ** * ******* to *** ****** ******** * ********, high ********** *****.

Vs ***** **** *** ******

*** **** **** **** *********** ****** performed ****** **** ******** **** ***************** ** *** ******* ** *** key ****:

  • ***** *** ***** ****** ******* *** person *** ******* ******* ***** **** two ***** ** *******, *** **** Lens **** *********** ****** *** *** mistake ****** *** ******* ****** ** ~1 ***** ** *******.
  • ************, *** ********* ****** ******** *********** under ~* *** *** ** *** integrated **, ***** ***** **** *********** camera *** *** *******.

*******, *** ***** **** ****** *** not ****** **** ********* ****** ********* issues ** *** *****, * ******* issue ** *** ********* *****. *******, the ********* **** **** ****** ** more ********* (~$*,*** **. ~$*,***) *** requires * ******** ***** ****** ******* of ***.

Vs *********

*********'* **** **** **** *** ****** has *** *** ********** *************:

  • *****, ****** *** ****** ** ********* compared ** ******* ** *******, ** is **** ***** ****** **** **** a ***** ********* **************, ***** ****** at $*,*** **** *** *******, *** does *** ******* ****** ******** ** cameras.
  • ******, *** **** **** ****** ********* much ****** ** *** *** ** scenes **** *********. **** **** ******* with **** ********** ****** **** **** and *****, *********'* ********** ******* *********** in ***** ******.

*******, ** *** *** *********, ** well:

  • ********* ******* **** ********* ****** ********* similar ** *********. *******, ** ******** no ***** ** ****** ***** ******* or ****** **** **** ****** ********** as ********* ****.
  • ******, *** ****** *** **** ** not ***** *** ** *** ******** search ******** ***** ** *********, **** as ********* ****** ** ** ******* in ******** *****, ****** *** ******** subjects, ** ****** ** ********* ****** cameras.

*******

*** ********* ***-**********-* *** ** *********** *** ~$*,*** ***.

Physical ********

*** **** **** **** *********** ****** looks ****** ********* **** *** ******* security ******, *** **** **** ***** two ******* ***** ** **** ********* to * **** *** ********** *** network *** ***** ***********. ***** ** review ***** ******** ***************.

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

*** **** **** ****** *** *** mistake *** ****** *** ******* ********** a ***** ** ******* *** ***+ people ** *** ********. ********** *** very **** ** **** ******, **** ~85-95%, ********* *** ***** (~*-* ***), dark (~*.** ***) **** **, ***, and ******* **** ****** *** *********.

Screen Shot 2019-09-24 at 7 24 21 PM

Short ********* ***** *** ******* ***********

***** ** ****** ******* *** ******, the ****** *********** *******, *********** *** optimal ** ******** ****** ******* **** ~6° ** ~**°. ** ***** **** missed ***** ***** ******** ** ****** lit ******.

*******, ********* ******** ** ******* ** ~10-11' (~**-*** ***) ****** *** ******'* 90° ***** ** ****. ******** ******* this ***** **** ****** *** ********, making *** ****** ******** ** ***** areas.

Screen Shot 2019-09-24 at 7 48 25 PM

Low ********** *** ****** ********* ** ****** *****

***** ***** ******* ***** ********* ******* in ****** ***** *********, *** *********** beyond ~**-**' *** ********, **** **** lower ********** ****** (~**-** **. **-**). However, **** ***** ***** ******** ** still *** ** ***** ************.

Confidence *******/****** ** ****** ********* ** **

** * **** ***** (~*.****), ********** drops ~** ****** **** *** **** is ******* ******* *** *** ****** has ******* ******** (~*°). ******* ******** of ~**° ******* ** ****** ********** or *** ********** ***********.

Poor *** ***** *********** (~* ***)

** * *** ***** *****, ***** were **** ******** ** * *** confidence (~**-**) **** **** * *** mounting ******/******** ***** *** ** ***** in *** ***** *********** *** **** of *** ****. ***** **** ****** detected ** ******* ****** ** *********, even **** ***** ******** ******.

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

*** **** **** **** *********** ****** was **** ** ******** ********* ****** at "******" ********** (~**-**) ** *** scenes, **** ** *** ******** ******* lobby **** ******* *****. ***** ******* and ****** *********** ******** ** **** tested **** ****** ** ******** ******* faces ** **** *****.

Lower ********** **** ****** *** ******* ***/**********

********** *** ********* ~** ****** ***** when ******** **** ******* ****/********** ******* through *** *****, *** *********** *** still ***** ** **** *** ******.

********* *** ***** **** ** *** face, **** ** * ***** ** mask, ******** ** * ***** ****, with ** ********* ** ***********, ********** of ********, *****, ** ********.

Inanimate ******* ******** ** *****

****** *******, *** **** **** ****** detected ******** ********* ******* ** *****. For *******, *** *** ** *** field ** **** ***** *** ******** constantly. ** **** *****, ********* ******* made ** * *********** ******* ** detections *** ***.

**** **** ***** ** ** *** to **** ******* ** *** * face ** ****** **** ** ****** detections ** *** ******** ***********.

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

*** **** **** ****** **** *** include ******** *********. ******* ******* *** scene **** * ******* ******* ** a *******'* **** ******** * **** confidence ******** ******.

Face ******** ******* ******

*** ****** ******** **** ********, ***** is ******** ** ****** ******** ** the ****** **** * **** ** detected. *******, **** ***** ** ** integrated ** ******, ******** **** *********** from ~**' ********, *********** ******** ******* and **** ********* ***********.

************, ******* ****** ** ***** **** exposure *** ******* ******** ** * totally ***** *****. **** ***** ******** for ~* ******* ***** ******** ********** properly. [[***, **** *** ********* *** about ****?]]

iVMS ***********

*** **** **** **** *********** ****** integrated **** ********** *** ****-****, ******** live ******* ** ***** ******** *** recognized ***** **** ****** *************.

Third ***** *** ***********

********* **** **** **** *********** ****** integrates **** ********* *** ************ **** *********** ******, ***** * **** *********** ***** is ***** ** ** **** **** in *** ******. *** ****** **** offers ** ***** *** ***** ** the ****** ******* ******* *** ***** when *** ***** ** *********, ** does *** ******* *** ****** ******** from *** ******* ** *** ******** metadata.

*** **** *********** ********* ** *** integrate **** ********, ***********, ******* ******** Center, ** ******* ***** *********.

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

**** ** ******** ** *** **** lens ****** **** ** *******, ******* the **** **** ** ****** *** camera **** ** ********** ** *********** and ******* ** *** ****** ********* North ******** ********.

*** *******, ***** *****, ** **** unable ** ***** ******* *****/******** *********** when ******** ***** ******* ********* *** in *******:

Setup/Adding ***** ** *******

******* ** **** *********** ** ******* to ***** ********* *********. ***** ****** need ** **** *** ********* *** exclusion ***** (** *******) *** ***** the "******* ***** ********" (*** ****** between ******). ** ******* ******* ***** distance ****** *****.

Single ** **** ******

***** *** ** ***** ** *** library ************ ** ** **** ** selecting * ****** **** ** ***** files. *****, ** *********** *** ******* of ****** ***** ** * ******* in **** *** ******** ** ***** recognition ** *** ******.

Country ** ******

*** **-**********-* *** ************ ** *****.

Versions ****

*** ********* ******** **** **** ****** testing:

  • **-**********-*: **.*.** ***** ******

Comments (15)

Does it integrate with access control also?

Agree
Disagree
Informative
Unhelpful
Funny

Great, now all our names & faces can be added to the PRC master database

Agree: 3
Disagree
Informative
Unhelpful
Funny: 3

Running out of products to test? For crying out loud, HIK IS A HUMAN RIGHTS ABUSER AND YET YOU CONTINUE TO PROMOTE THEM! Unbelievable.

Agree: 2
Disagree: 4
Informative
Unhelpful
Funny: 2

AND YET YOU CONTINUE TO PROMOTE THEM!

#3, thanks for your feedback. I've addressed this at length here - Should IPVM Ban Testing Dahua And Hikvision Products Because Of Human Rights Abuses?

Short version - doing in-depth, independent, analytic tests is not promotion and is critical to delivering accurate reporting.

Agree: 2
Disagree
Informative
Unhelpful
Funny

John, I feel sorry for you

You are given a hard time for your admitted & honest anti hikau feelings.

You are given a hard time because despite the above you ensure honest, factual reviews of these products.

You seem to be upsetting both pro & anti hikau camps, this means you are almost certainly being honest & objective with your reports.

At the end of the day this is a technical resource website with unbiased reports on the capabilities & performance of IP CCTV devices.

Agree: 2
Disagree
Informative
Unhelpful
Funny

What about Apple, or Microsoft? Both have been accused of Human Rights violations. Should we stop using our iPhones and Windows based computers?

Or Amazon? With numerous reports of not even letting their distribution warehouse workers use the bathroom, or penalizing them for using the washroom?

I am not denying, or defending what HikVision has done, but why isolate one, and not the others? I get this is a site dedicated to Security stuff, but heck most VMS and Access Control software runs on Windows. Many of us use and operate Apple products as part of both work and use of Security. Many people probably buy all sorts of products used for Security off Amazon. I've bought tools there.

Where do you draw the line? Why only HikVision? Compared to Microsoft, Google, Apple, or Amazon, they are tiny. Owned by a communist government, sure. But all have done, and continue to do, bad things.

Agree: 1
Disagree: 1
Informative
Unhelpful
Funny

Where do you draw the line? Why only HikVision? Compared to Microsoft, Google, Apple, or Amazon, they are tiny. Owned by a communist government, sure. But all have done, and continue to do, bad things.

Daniel, when was Google or Apple found by any government (US or PRC or otherwise), to commit human rights abuses?

My concern is that your tactic can be used to absolve any organization of anything. Dahua has been paid hundreds of millions of dollars to build and operate police stations in Xinjiang that are used to oppress Uyghurs. Does that not cross 'the line' for you?

Agree
Disagree
Informative
Unhelpful
Funny

I guess I should clarify, by "Tiny" I mean in terms of physical size, revenues, value, etc.

According to this newsweek article, Apple has had several Human and Labour Rights violations in it's Global Supply Chain, with the number of violations growing. Not as big as what HikVision is doing, no, but still violations.

Marco Rubio has accused Microsoft of being Complicit with what China is doing.

Amnesty Internation as accused numerous companies of Child Labour. Now I can easily admit that's a bit of a stretch, as Apple, Microsoft, Google, Samsung, whatever, do not do the mining of the material they use. That's really on the mining companies.

Are any of these examples as bad as what HikVision is doing? I'm not trying to imply that because other companies do bad things that we should ignore what HikVision is doing.

It's good people talk about that issue and re-consider using the product. But why stop there? Is the point I was trying to make. So UM3 doesn't want IPVM to test HikVision products. Why stop there? HikVision bad, yes. Dahua also bad for your linked example. I don't disagree with that one bit. But what about the other companies committing lesser violations? Why only single out Hik or Dahua, and not others involved?

Agree
Disagree: 1
Informative
Unhelpful
Funny

Same reason why murder carries a longer prison sentence than jay walking.

Agree
Disagree
Informative
Unhelpful
Funny

Oh, hi Dahua...

Agree
Disagree
Informative
Unhelpful
Funny

I might have overlooked it in the article, but how many people were in the Gallery during the test period?

Agree
Disagree
Informative
Unhelpful
Funny

We had a database of 100 faces during testing and some additional faces were added from unknown people walking through which brought it to 108.

Agree
Disagree
Informative
Unhelpful
Funny

8 acts of privacy violations =)

Agree
Disagree
Informative
Unhelpful
Funny

It would be interesting to know or test how many faces can be processed simultaneously.

Agree: 1
Disagree
Informative
Unhelpful
Funny

Big yes on this. But I am guessing it takes alot of processing power, which is why Anyvision is so costly.

Agree
Disagree
Informative
Unhelpful
Funny
Subscribe to IPVM Research to read the full report.
Why do I need to subscribe?
The IPVM Research Service includes products tests and shootouts plus competitive and financial analysis, helping decision-makers better evaluate purchasing, partnering, developing, and/or competing against companies in physical security.
Already have an account?
Loading Related Reports