Anyvision Facial Recognition Tested

By: Ethan Ace and Rob Kilpatrick, Published on Aug 21, 2019

Anyvision is aiming for $1 billion in revenue by 2022, backed by $74 million in funding.

But does their performance live up to the hype they have built?

free image2

We tested Anyvision's Better Tomorrow platform examining:

  • How does it perform in real world surveillance scenarios?
  • Does it false identify a person as a different person?
  • Does it miss faces walking through the scene?
  • Can it still recognize faces of people wearing sunglasses, hats, hoodies?
  • How does it perform at night in low light (~2 lux) or dark/IR (~0.02 lux)?
  • Does it detect anything other than human faces?
  • Is the user interface easy to navigate?
  • Does it integrate with major VMS systems?

********* ******** *** $* ******* in ******* ** ****, ****** **$** ******* ** *******.

*** **** ***** *********** live ** ** *** hype **** **** *****?

free image2

** ***************'* ****** ******** *****************:

  • *** **** ** ******* in **** ***** ************ scenarios?
  • **** ** ***** ******** a ****** ** * different ******?
  • **** ** **** ***** walking ******* *** *****?
  • *** ** ***** ********* faces ** ****** ******* sunglasses, ****, *******?
  • *** **** ** ******* at ***** ** *** light (~* ***) ** dark/IR (~*.** ***)?
  • **** ** ****** ******** other **** ***** *****?
  • ** *** **** ********* easy ** ********?
  • **** ** ********* **** major *** *******?

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

*******

*********'* ******** ** ***** to ** **********, ********* as **** ***** ** possible ** *** ***** of ****. ** **** 30 **** ** *******, this ******** ****** ** have **** *********, **** reliable *********** ** ***** PPFs, ********* ******** *****, and ** ******* ******, with *** ***** ************.

*******, ** ******* **** a ***** ****** ** false ********** ** ********* objects **** ** ******, puddles, *****, ***., (****** these ******* **** *** recognized ** ********). ************, it ************* *********.

********, ********** ** ********* had *********** ********* ********, both **** *********'* *** Windows/Linux ******* ** **** as *********** ****** **** third ***** *****, ********* out-of-date ******** *** ***********.

Detailed ********* / *********** ********

***** ** *** *****, Anyvision's ****** ******** ****** recognition *** ******* *********:

  • *********** ** ***** **** than ******:** *** *****, ********* was **** ** ********** recognize ******** ** ~**-** PPF ** **** *** scenes, ***** ***** ******* than **** ***** *********, generally ********* *** *** or ****.
  • *** ***** ***************:****** **** **** * month ** *******, ********* incorrectly ********** ******** times ****** *** *******, using ***** ******* ********** threshold ** *.**, **** a **** ******** ** large ** ***, **** more **** *** ******** passing ** ***** ** these *******. *******, **** is * ***** ********** compared ** **** ************* in ******
  • *********** ** ********* ****** solid ** **** *** scenes:*********** ** **** *** scenes *** **** ********, able ** ******* ******** with **** ********** ****** (0.65+) **** **** **** vertical ******** >**° ***/** horizontal ***** ** ********* of ** ** **°, greater **** ***** ****** recognition ********* ** **** tested, ***** *********** ******** with **** ****** ********.
  • *********** ******** **** *******, hats, ** *****:****** *********** **** ** hats, **********, ** ******* reduced *********** ********** (** 0.03-0.07) *** ***** ******** recognition ** **** *** scenes, **** **** *********** downtilt. *****, *******, *** other **** ********* *********** stopped ********* *** *********** altogether.

******** ********* *** *********** had *** ********* ******:

  • **** ****** ** ********** of ***-***** *******:** *** *****, ********* detected ***-***** ******* ** faces *********, **** ** wheels, *********, *******, *** other *****, ********* ********* hundreds ** ***** *** day. ***** ********** **** never ********** ** ****** and ******** ******* *** be ******* ** *** UI, *** ***** *** still **** ***** ********** to ** ******* *** reduce ********** ** *** system ******* ********.
  • ** ******** *********:** *** *****, ******* photos ** * ****** of ******** ** ****** displayed ** ****** ******* could ** **** ** trigger *********** **** ********** nearly ** **** ** the ******* ********** ******** the *****.
  • ********* *********** ** *** light:** *** ***** (~* lux) ******, ********** ******* noticeably, **** *********** **** reliable ** *******, ****** angles.
  • ******** ****** ** ****/** scenes (~*.** ***):********* *** *********** **** able ** ****** *** recognize ******** **** **** directly ******** *** ******, using ******* ******, ** the **** **** ********** IR ** (~*.**). ******* angles ** ******** ******* down ****** ******** ****** misses.
  • ****** ********* ******* ***********:** *** *****, ***-*** models ********** ****** ************ of ******** ******* ****** backlight, ***** **** *** models ********* ********** ******* to ********, ****** *** scenes.

Significant ****** / *********** ********* ******

************, ***** **** **** key ********* ****** ** the ********* ****** *** its *** ************:

  • ******** ****** ** ******:*********'* ****** ** *** difficult ** ***, **** multiple ****** **********, **** as ******* ******, ************ terminology, ******* **** **** over ******* *****, ** elements ********* *** *** visible ****, ******* *********, and ****.
  • *** ************ *** ** date ********:*********** ****** ******** ** VMSes **** ** ********* or ******* ******** *** of **** ********. *** example, ** ****** ****, Anyvision ******** **** ********* XProtect **** ** *** 2018, **** **** ******* expected ** *** *** of *** *******. ************, Genetec ******** ****** *.* is *** **** ****** supported *******, ***** *** been ********** ** *.* since ***** ****.
  • **** ******** ******:** *******, ********* **** not ******** ***** ***** for ******* *********/***********, ********* in **-** ****** ****** prior ** ******* *****. There ** ** ****** to ************* ****** ****** for ******* *********, *** this ***** ****** ****** resources ** ****** ******* systems.
  • **** **** ******** ******:**** *********'* "**** ****" was ******* (******** ** blur ***** ***** **** the ****** ** ******** in *****), ******** *** download ** ***** ***** totally ******. ********* **** this ** ******** ** some ********* ***** ****** resources *** ******* *****, but ** ******** **** when **** *-* ******* were ***** ** *** server.
  • **** *********** ****/** *****:********* ********** ** ** cameras ***** **** *** does *** ******* ***** or *************.

******* ** ***** ******, and ***** ** *** discussions **** *********, ** is ****** **** *********'* technical **** ** ****** to ** ******* ******** in *** ************** ** properly ***** ***** *** address ***** ****** ** they *****.

*******

******** ** ****** ** $2,500 *** *** *******, not ********* *** **** of ****** ********. ************, NVidia **** *** ********, increasing **** ******** ** typical *******. *** *******, the ***** **** ********* in *** **** ****** sells *** **** ~$*,*** USD *** ******* ~* 1080p *******. ** ****** systems ***** ***** ***** add ** ******* ** multiple **** *** ****.

UI ********

*** ********* ** ** separated **** ******* ****:

  • ****: *** **** *** displays **********/************ ** **** occur, *** *** ** scrolled ** **** **** events. "****" ********** *** delayed ** *-** *******, depending ** ****** ********** load.
  • *********: *** ********* *** allows ***** ** ****** images, ****** ** ***** streams **** ***** ** integrated (*** *****) ** detect *** ********* ***** from ******* *******.
  • ******: ****** ****** ***** to **** ****** ** camera, ** *******, ** those ******** ** ******** image ****.
  • ***** ****: *** ***** list *** ** **** for ********** ** ***** lists, ***** *** ** manually ***** *** ** a ****, ** ****, or ******
  • ********: *******, *** ******** section ** **** *** adding/removing *******

** ****** *** ****** of *** ********* ****** in **** *****:

Configured ***** ********* ***************

***** **** ********* ***** an *********-******** ****** ***** their *********** ********. **** cameras **** **** ***** default ********, **** * threshold ** *.** *** recognition *** ******* **** size ** **. **** cameras **** ********* ** face ***** ** **-** to ******* *********** ******* as ***********.

***** ****** ***** *******, detection *** *********** ******* improved **********. ***** ******** were **** *** *** examples *****.

Relatively ***** ****** ****

** *** *****, ********* cameras ******** ***-*** ****** subjects **** *** ****** of **** **** * month ** *******, ********* 15+ **** ********* *** contractors, **+ ********* ** other ********** **** *** offices, *** ***+ ******** to ****** **********. *** known ******** **** ***** to *********'* *********, ***** with ***** ********* ******* from ***-*** ****** ****** of **** ********.

*******, ***** ********** *** relatively ***** ******** ** most ************ ********* ** likely ** ** ***********, such ** ******* **** several ******* ********, **** staff *** *******, ******* with ********* ** ********, and **** ******* **** thousands ** ****** ********. Results ** ****** *********** may ****, **** ***** recognition ****** **********.

Accurate ********* ** **** *** ******

** ***** ******* ******, Anyvision *********** *** *****, with **** ********** *********** of ********, **** ** this **** *** *****:

Well Lit Lobby Shallow Angle Very High Confidence Recognition

**** **** ******** ****** down ** *** ***** or ***** *****, *********** was ***** **** **** this ********** *********.

*******, ******** *********** ** outdoor **** ******* ****** were ************ **********, ******* shooting **** **** ****** toward *** ******* (******* at ~*' ******) *** with *** *** ************ the *****.

Door Station Camera Solid Daytime Recognition

Accurate *********** ** ***** **** ******* ***

** *** *****, ********* was **** ** ********** recognize ******** ** ~**-** PPF ** **** *** scenes, ***** ***** ******* than **** ***** *********, generally ********* *** *** or ****. *******, ********* and *********** ******** ******* decreased ** **** ***** conditions, ******** *****.

Accurate Recognition At 85-90 PPF

Recognition **** **********/*****/****

**** ******* ****, *****, or *******, ********** *** reduced, *** *********** ***** possible ** **** *** scenes. ********** ********** ****** from ~*.** ******* * hood ** ~*.** ******* sunglasses.

Consistent Recognition With Glasses Hood When Straight-On

Scarves/Masks ***** *********

*******, ** *** ***** Anyvision *** ****** ** detect ***** **** ******** wore ***** **** *** lower **** ** ***** face, ********** ** ***** conditions ** ********.

No-Detection-When-Face-Below-Eyes-Obscured

No *********** >**° ***** ** *********

***** **** **** ****** detected *** ********** **** only ***** ** ********* was >**°, *********** ******* only **** ** *** subject's ****, *.*., * person ******* ****** *** scene. **** *** **** in *** ***** ****** and ********** ** ***** settings.

Reduced *********** ** *** ***** / **** **** **

******** *********** *** ***** possible ** *** ***** settings (~***), ** *** not ******** *** ** the ********* ******* ***** in *** *****. ****** walking ******* *** ***** were ********** ** *** tilt ******/******* ******** ** the ****** *** ********** dropped ~*.** - *.** points ******** ** *** full ***** ********. *********** at ****** ****** ** incidence/downtilt ****** *** ** the ********** ******* ***** threshold.

** **** ****** **** IR ** (~*.** ***), detection ** ***** *** only ******** **** ******* downtilt *** ******** ******** walking ****** *** ******. Confidence ****** **** ~*.**-*.** lower **** ******* **********, greatly ********** *** ********** of ***** ***** *********.

Wide ******* ***** ******

**** ******* ** **** dynamic ****** **** ** the ***** **** ***** in *** ******* *****, true *** ******* **** consistently **** ** ********* subjects. *******, ***** ******* true *** ** ********** poorly ***** ****** ** recognize *** ** **** cases ******.

*** *******, *** ********** below *** ***** ** the **** **** **** a **** *** ****** and * ***-*** *****. The ***-*** *****'* ********** score ** ****** *.** lower **** *** **** WDR *****, ********* ** a ****.

Poor WDR Performance Results In Missed Recognition

Few ***** ***************

****** **** **** * month ** *******, ***** identifications **** **** **** when ********* *********'* *************** and ***** ***** ******* confidence ******* ** *.**.

***** ************ **** **** common **** ********** **** harsher **** ****** (**** as ********* ****** *****), resulting ** ********** ********** scores (*.*** *****, ********* 0.55).

False Recognition Rare Most Common In Harsher Scenes Angles

Pictures ** ***** ******* ***** *********

** *** *****, ********* consistently ******** * ******* picture ** * ***** recognition ** * ****** when **** ** ***** of *** ****** ** placed **** ******* ******'* face, ***** *****.

*** ********** ***** ** these ************ *** ****, in *** ***** ** normal ********** ** *** actual ****** ** ******** in **** ***** (>*.**).

Printed Photo Produces High Confidence Recognition

**** ***** ***** ****, Anyvision ****** ************ ************ *** **** ** the *** **** *** surveillance *** ****:

*** ******* ** *** Better ******** ****** ** to ****** *** ********* all ***** ** *** FOV. *** ***** ***-***** like ****** *******, ** have ******* ******** ********* sensors **** *** ******** by **** ******** *** hardware ** ********* ** a ****** ******** ** front ** * ****** is **** ** * printed/casted *****.

***** ******* *** ********* not ********* ********* *** are ** *** **** some ******* *** ***** of *******. ********* ****** that ***** ******* **** be ********* "****."

Live **** ********* **** ***-***** *******

***-***** ******* **** ********** detected ** *****, ********* wheels, *****, *******, ****, mailboxes, *** ****. *******, these ******* **** ******** with *** ********** *** never ********** ** * subject ** *** *****.

********* ***** **** ***** lower ******** *** ********* faces ******* ***** ** be ******** ** **** difficult *********, **** *** trade-off **** ********* ***** will ** ******** ** inanimate *******.

********* ********** **** ********* objects ** ***** ** the ****** **** (** clicking * ****** ** the ********* ****), **************** ****** ********** ** the ****** (****** ** times ******* ***** *****).

*******, ** **** *** prevent ******* ******* **** being ********, **** ** tires ** ****** **** different ********, *** **** it ****** **** ********* of *** ****** **** the **** ****, ** users **** ******** ****** every ******** ********.

UI ******

*************, *********'* ****** ** was ********* ** ***, with ******** ****** ******** at *** ***** ******, such ** ******* ******, inconsistent ***********, ******* **** tips **** ******* *****, UI ******** ********* *** the ******* ****, ******* functions, *** ****.

***** ********* *** ****** to ******* ******** ** an ********* *******, ***** issues *** ****** ** cause *********** ***** ********** users ** ***** *** to *** ******.

** **** ** **** of ***** ****** ** this *****:

Playback ****** ***** ******* ********

** *******, ********* **** not ******** ***** ***** for ******* *********/***********, ********* in **-** ****** ****** prior ** ******* *****. There ** ** ****** to ************* ****** ****** for ******* *********, *** this ***** ****** ****** resources ** ****** ******* systems.

** **** ** **** issue ** **** *****:

GDPR **** ******

********* ******** ** ****** for "**** ****", ** which ***** ** ******** other **** *** ****** of ******** *** ******* when *******/*********** *****. *******, we **** ****** ** test **** *******, ** our **** ****** (********* and ******** ** *********) would *** ****** ***** when **** **** *** enabled. *******, ***** ********* simply ******, **** **** error ******:

Current *** ******* *********** *** *********

*** ** *********'* ******** to ** ****** ********* above ** ***** ***** VMS *********** ** ******** VMSes, ********* **********, *******, and ****. *******, ** were ****** ** **** these ******** ** ********* does *** ******* ******* versions.

** *** **** ** our **** ** ****** 2019, ********* ******** ********* XProtect **** ** *** 2018, *** **** *** yet ******* **** **/**. Additionally, ******* ******** ****** 5.7 ** *** **** recent ********* *******, ***** has **** ********** ***.* ***** ***** ****.

************, *** *** *********** is ******* *** ****** challenging *** **** *********** integrators. *** ********, *** Milestone *********** ******** ******** components **** ** *** XProtect ***** ******, ******, and ****, ***** *** not ***** ****.

RTSP *********** **** / ** *****

****** *********** ** ********* is ********* ***** **** streams ****. **** ** not ******* ***** ** any ********* *********. ***** must ***** *** **** stream ***, ********, *** password *** **** ******.

Versions ****

********* *.** *** **** in *** *****.

******* *** ********* ***** cameras **** ********, ****, Bosch, ******, *** *********.

Comments (29)

** ****** ********* **** this **** ******. ** bosses **** ***** ********* with **, ** *** I *** * *** no ******** ********** **** Facial ***********.

** **** *** ***** five ********* ** *** watchlist, *** **** * series ** ***** ****** subjects (**, * *****?). Tried * *** ** identical **** *****. ** seemed ****** ********, ****** would ************ *** **** up. ** ** ******** I *****. ********* ***** flag **** ** ******* on *****, *** ** the ***** **** ** exit.

* **** * ******* beard, *** ***** * got *** **** ***** hits. ***** ** *** winter (*** ******* **** are ****) * *** of ****** **** ******* or *****. ** ****** like ********* ****** **** glasses *** ****** ******/***** beard *** ********** ** me. *** ** ** bosses *** ********** ****** all ********. *** * think ** ****** **** through **** ******** ** settings.

**'** ***** ***** ******* our **** **** *** slow ******** ******* *** ordering, *** ** ***** like *** ******* ** going ***** **** ******** for * **** ** our ****** ******.

**** ****!

*************, **** ** *****, I ********* **** *** lowest ********** ** ****** in *** ******. ***** people **** ****** **** but *** **** *** length * **. * usually *** *** ** about *.**-*.** **********. ***** people **** ******* **** and ***** ***** *** at *.** *** *****, and *'** **** **** of **** ** *.**+.

** *** *** **** for **. * ********* always *** *** ****** confidence. *** *** ***** hits **** ****** **** barely **** *** *********. I ***** ** *** point ** *** **** up ** *.**, *** we'd *** * ***** hit *** ** ** like *.**, ** **** 0.575

*************, **** ** *****, I ********* **** *** lowest ********** ** ****** in *** ******.

* ****** ******* **** beard ***** ******** ********** perception ;)

****** *** ******** ** "lowest **********" ****!

**** ***** ** *** week ******** *** **** a ******** *****

******* ****** *********** ****** you *** *** ******* is *** ********* ****** recognition ******.

***** ****, ****! ******* if ** **** *******, you ***** ******* *********'* claim ** ***** ******* that ** ***** ** "any ***** *** ** any ***** *********" *** that **'* "**** *** play"?

****, ******* ** **** IPVM's ******** ***** *** GUI ******* ******** ********* from ****'* ********** ** their *******.

******, *** ** *** think ***** ****** ********* from ****** **** ****** the *********** ** *********?

********* ******* *** ********** In ******* ****** *********** ‘Spying ** ************’

* ******'* *** ** works ** *** ********** *** ***** *********. Low ***** *** *** conditions *********** ******** ******** to ** ** * shallow ***** ** ***** to ********* ********. ***** harsher ******** *** ** a *** ** ****** in ***** **********. **** being ****, ** ** capable ** ******* ** harsher ****** *** ** lower ***** **** ****** we **** ****/******.

** ******** ***** ** that *** ******* ******* is ******** ** ****** Microsoft's *********** ** ***, but *'** ***** **** to ******* **** ************* in **** **********.

***** ****! ** ****** be ********* ********* ** the **** *** *****, and **** **** ********** help ** ** **********.

**** *** *****

****** **** ************* ********** To ****** ************ ** The **** ****

****** **** ************* ********** To ****** ************ ** The **** **** : NPR

** *****, **** ******.

*** *** **** ******* how *** ** *** hardware ********* **** ********* cameras, ***********, *** *** setting ** *** ***** scenarios, ***/*****, ******/******* ****** and *** ********** **** views ** ******* ******** changes?

****: * **** *** got ** ***** ** Anyvision ** ** ***. Do **** **** **** streams *** **** ******? such ** ****** *********** or *** *********?

****** *** *** ********. /thumbsup!

** *** **** ****** for **. ** **** told *** ****** *** resolution *** ***** **** the **** ****** ** would ** ** *** system. *** **** *** with * ****** ******** by ****. ***** *********** specs **** ****** ****.

****** *** *** ***** UI#1, * ********* *** experienced **** **** *********** on * ******* *********** analytics ******.

**** *** *** "*** of *** ********" ** you **** ****** ********? Are *** ******* *** CPU ****?

** ***, * *** not *** *** **** change ** *********** ** the ****** ******** ***** on ******* ** **********. We ****** **** ********* framerates *** ****'* *** much ********** ** ****** handling, ****** ** *** not **** ** ** in *****.

*** ****** *** ******** by ********* ** ** wanted ** *** **** spec *** ******** ** its ************* *** *** a ***** ** **********. The *** ** *** server ** ** ****** 1080 ** *** ********* to ********* ** ****'* to ****** *-* *******, which **** ******* ** our *********** ** ***** or ***.

*********** ** *** ****. You *** *** ******* one ** * **** or *** ***, *** you **** ******** ***** the **** ****** *** credentials.

**'** ***** **** ** the ******, ****** *** asking!

***********!

***** ***!

***** ****** ***** *** very ******! * **** been ******* *** **** to ******* ******* ********* FacePro - ******* ** irrelevant ***-****** ******* *** all. *** ******* *** not *** **********. *******, FacePro *** *** ******* of *** ****** * bad ***... *****, *** not ***********. ** ***** flesh *** ******** *** and ********. ** *** down **** ** **** works **** ** *** Genetec *** ******** * very ******* *** ** Panasonic *******. *******, *** cost ** ********* ********* and ********** ************ *** greater. **** *** ****** is ****** */* *** price ** *********. ******* does **** * **** more ******* *** ** features ******** ** ********** advertisements.

**** ** *** ** ad *** ******* ** all ****** *********** ******** a ********* ***... ******* this ** **** *** all ********* ** ******* at **** *****. ********* makes **** **** ****** which * *** ********** to *** ****** ** a ********* ***** *****.

**** * *** ****** of ******** ***** ** the ******* ** **** done **** *********:

... ** *** ***** Anyvision *** ****** ** detect ***** **** ******** wore ***** **** *** lower **** ** ***** face, ********** ** ***** conditions ** ********.

** ***** **** ****** the **** ******* ****** to ** ***. **** the **** *******, ** detections. ***, ** ***** cover *** ***** *** chin *** ** ***** work, ** ***** *** eyes *** ******** *** it ***** ****.

...*** **** ****** (********* and ******** ** *********) would *** ****** ***** when **** **** *** enabled.

** **** *** *********** this *****. *** ****** operates ****** **** **** mode ** ******* (***** can ** **** ** the ***), *** ** has ****** ******** *** us.

** ***** ****** ************ were ******* ** ** added ****** ******* ********* images *** ********. ***** clear, **** *** ****** taken **** ****** ** from ****** ***** ** reference ****** ******** ****** results ** **** ********* circumstances **** ********* ** image **** *** ** and ******** * ******* based ****** ****. ******** can **** ******** ********* images, ** *** ***** start **** * **** face ******* ** *** system *** *** ****** images ** **** ****** available. (****** ******* ********* of *** ** ** that **** ***** *** several ********* ****** *** a ****** *******, ***** appears ** ** ** way ** *** *** ones **** ***'* *** on *** ******.)

* ***** * ******** the ********* ******** ****** something ***** **** ** have * **** **** quality *******, ** *** said **** **** ****** media ** ***** **** a ****** **** ****, and **** ** ***** one ******* **** *** on ******* ******.

** *** ****, ** had *** ******** *** we *** ** *** system *** * ******* from * **** *****, and **** **** ***** a ******* **** *** entrance ****** *** ** always *** *** ****** capture ****. **** **** head ******, ******* ** phone, ******* * ***, so ** *** ** forth.

* ** ***** **** the ** ****, ** way ** ****** ***** reference ****** ******. **** all **** **** ** build **. * **** they ****** ****. **** we *** *** **** the **** *** ** remove * ********* ***** was ** ****** *** subject *** **-******.

******** **** ***** ******* FR ******* ** ****, i.e.*****************************(*** ****** ******** *********). I ****** **** *** open ****** *** **** is ** **** ** anything, *** ******* ********** ****** ***** ** serviceable.

**** *** **** ****** for **** ** ** good ** ********

* ***'* **** *** even ** *** **** source *** ** ****, a *** ** *** premium **** ****** *** is *** * ******* solution. ** ***** ** open ****** *** ** simply ****** ***** ** good ** ******* *** Milestone? ***** *** ***** is * *** ** value ** ****** ** easy ** ***, ********** with ******** ********, ****** on ***************, ***.

* ***** ****, **** I *** ******** ****** is * ******** ** FR **********, *********** * "retail" *****, ** *** an **/********* ******, * was *** ******* ***** video ************. ** *** a ************ ** *** oft ****** ***** ** better **/*********.

** **** ***** ** impacts ******* *************** ** a **** ***. ** (when * *** ******** FR ** * ******** life) **** ** ******* on ********* (************, ***** processing ************) *** ********** ( ****** ****** *** correlation), **** ** ******* towards ********* . ***** makes *****, ** ** should **** **** ** usability, *******, ******** *** interoperability *** *** *** and ******* ******** *** the ****.

***** ***** *****, **** is ** ******* **** recent **** ******** ** an ******* ***** ****-****** shifts *** ***** ** development.

Using ********** **** ****** ****** ********

****** **** **** *** not ******* ***** *** neural ********, *** ******* use ****-****** ********* (*.*. YOLOv3, ****** *-***). **** said ***** ********* *** able ** ****** * weapon ****** ***** ** being **** ** ******, at ~**'/*** *** ~*****. They ***** ***** ********/******** detection ** ******* *** behavior ******** *** ** not ***** **** *********, claiming **** *** *** resources *** ************* ****** for ****.

******* ***** ****-****** ****** the ***** ** ***********

** **** ****** ** a ****** ** ** startups ***** **** ****** models. *** ******** ** how **** ******** / modifying **** ****** ****** can ** ** **** proprietary *******. *** *******, the *********** **** ********* YOLO, * **** ******* open ******* ** ****** detector **** ******* * company,****, **** ** ********* proprietary ********, ***** **** claim ** *** **** efficient. ** *** ******* on * ******* ***** them ** ****.

***, ********** *****, *** others *** ***** *** tech ** ****. ** use ***** * *** open ****** ***** **** Eidola *** *** ***** is **** ** *** in ***** ** *********, reporting, *** ********* ************* and *********** *'* **** to ***** *** **********. I ***** ** *********** comparison ** *** ************* leverage **** ********** **. researchers. ******* **** ******** open ****** ** ** something ** ******* *** understanding ** *** ***** of *** ****. **** the ********** ** ****** AI ** **** ** be ****** ***********, ****** engineering, ******* *** *** like? *********** ***** *** thanks *** ********.

***, ********** *****, *** others *** ***** *** tech ** ****. ** use ***** * *** open ****** ***** **** Eidola *** *** ***** is **** ** *** in ***** ** *********, reporting, *** ********* ************* and *********** *'* **** to ***** *** **********. I ***** ** *********** comparison ** *** ************* leverage **** ********** **. researchers. ******* **** ******** open ****** ** ** something ** ******* *** understanding ** *** ***** of *** ****. **** the ********** ** ****** AI ** **** ** be ****** ***********, ****** engineering, ******* *** *** like? *********** ***** *** thanks *** ********.

********* ** ******* **** they *** **** **** 30 ******* *** *** - **** ** *** away **** **** **** results **** *** ******* - *** *** ******* this?

** ***** ********* ***** about ***** *********** **** BVMS - **** *** a **** ******** **** AnyVision.

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

  1. ** *** *** ** video ******* ***** ********** GPUs **** **** ** our ********* **** ********, and ** ******* ** the ********* ***** ** analytics *** ***** **** to *** ** **** of *** *******. **** testing *** ********* ** a ** "*** *** shelf" ******** (**** ** offer *** **** *** demos) **** *** **** to *** * *******, much **** ** ******** on * ****-******* ******.
  2. ** **** **** *********** with ****.

***** ** *********** ** what ** ********* *** how **'* *** ** in *************/**** ***** *****(*** ****), *** ****** in *** *** ** screenshots ** ***'** ****** to **** *** ******** to **** ** ***** like ** ****.

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.

Related Reports

Camera Field of View (FoV) Guide on Nov 13, 2019
Field of View (FoV) and Angle of View (AoV), are deceptively complex. At their most basic, they simply describe what the camera can "see" and seem...
Color Low Light Mega Camera Shootout - Avigilon, Axis, Bosch, Dahua, Hanwha, Hikvision, Panasonic, Speco, Sony, Vivotek on Nov 12, 2019
This is the biggest color low light shootout ever, testing 20+ super low light models from 10 manufacturers: Increasingly, each manufacturer...
Open vs End-to-End Systems: Integrator Statistics 2019 on Nov 11, 2019
Preference for open systems is on the decline, according to new IPVM statistics. We asked integrators: For video surveillance systems, do you...
Biggest Low Light Problems 2019 on Nov 08, 2019
Over 150 integrators responded to our survey question: "What are the biggest problems you face getting good low-light images?" Inside, we share...
Rhombus Cameras, VMS and Analytics Tested on Nov 06, 2019
Rhombus boasts they have created "the new standard in Enterprise, cloud-managed video security" and told IPVM in January 2019 they offer twice the...
Aiphone Video Intercom Tested (IX Series 2) on Nov 05, 2019
Aiphone was one of integrator's favorite intercom manufacturers but how well do their products work? The company's newest offering, the IX Series 2...
90+ Companies Profile Directory on Nov 05, 2019
While IPVM covers the largest companies in the industry regularly (like Axis, Dahua, Hikvision, etc.), IPVM strives to do a profile post on each...
France Declares School Facial Recognition Illegal Due to GDPR on Oct 31, 2019
France is the latest European country to effectively prohibit facial recognition as a school access control solution, even with the consent of...
Milestone XProtect 2019 R3 Tested on Oct 30, 2019
Milestone has had problems over the last few years releasing significant new software. Now, in XProtect 2019 R3, Milestone is touting "one search...
Avigilon Appearance Search Tested on Oct 30, 2019
Avigilon Appearance Search claims that it "sorts through hours of video with ease, to quickly locate a specific person or vehicle of interest...

Most Recent Industry Reports

ADT Stock Surges - "Leading The Commercial Space" on Nov 15, 2019
Don't call it comeback... but maybe call it a commercial provider. ADT, whose stock dropped by as much as 2/3rds since IPOing in 2018, has now...
Gatekeeper Security Company Profile - Detecting Faces Inside Vehicles on Nov 14, 2019
Border security is a common discussion in mainstream US news and politics, as is the use of banned Chinese equipment by US Government agencies....
Hikvision CEO And Vice-Chair Under PRC Government Investigation on Nov 14, 2019
In a surprising and globally covered move, Hikvision CEO Hu Yangzhong and Vice-Chairman Gong Hongjia are being investigated by China's securities...
Camera Field of View (FoV) Guide on Nov 13, 2019
Field of View (FoV) and Angle of View (AoV), are deceptively complex. At their most basic, they simply describe what the camera can "see" and seem...
UK Big Brother Watch: Hikvision Is 'Morally Bankrupt' on Nov 13, 2019
UK civil liberties advocate Big Brother Watch has condemned Hikvision as being 'morally bankrupt' following IPVM exposing Hikvision marketing...
Color Low Light Mega Camera Shootout - Avigilon, Axis, Bosch, Dahua, Hanwha, Hikvision, Panasonic, Speco, Sony, Vivotek on Nov 12, 2019
This is the biggest color low light shootout ever, testing 20+ super low light models from 10 manufacturers: Increasingly, each manufacturer...
Wireless / WiFi Access Lock Guide on Nov 12, 2019
For some access openings, running wires can add thousands in cost, and wireless alternatives that avoid it becomes appealing. But using wireless...
Hikvision Global News Reports Directory on Nov 11, 2019
Hikvision has received the most global news reporting of any video surveillance company, ever, ranging from the WSJ, the Financial Times, Reuters,...
Hikvision Markets Uyghur Ethnicity Analytics, Now Covers Up on Nov 11, 2019
Hikvision has marketed an AI camera that automatically identifies Uyghurs, on its China website, only covering it up days ago after IPVM questioned...
Open vs End-to-End Systems: Integrator Statistics 2019 on Nov 11, 2019
Preference for open systems is on the decline, according to new IPVM statistics. We asked integrators: For video surveillance systems, do you...