Facial Recognition 101

By IPVM Team, Published Mar 18, 2020, 10:52am EDT

Facial recognition interest, use and fear is increasing. This guide aims to teach you the fundamentals of facial recognition.

IPVM Image

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

  • **** *********
  • **** ***********
  • *:* ** *:*
  • *********** ** *************
  • ********** / ***** ************
  • ********** ****** ******* ** Facial ***********
  • ******** / *********** ****** Recognition - *****, ****, Sunglasses, ***.
  • ******* **** ** ***** Face ***********
  • ****** / ****-******* *********** Issues
  • ******** *********
  • ********
  • ******* ********
  • **** *********** *********

Facial *********

****** **** *********** *** be *********, * **** needs ** ** ********. Typically, * *********** ****** is ***** **** *** face. **** ** * relatively ****** ********, *** much ****** *** **** accuracy **** **** ***********.

*********** ****** ** * fundamental *******:

  • ***** ** *****: ***** it ** '****' ** detect * **** ******* directly ** *** ******, performance *** **** ************* depending ** *** * person ***** ***** **** (down, ****, *****, ***.)
  • ******** ** ***** - While ** ** '****' to ****** * **** looking ******** ** *** camera ** * ****-*** scene, *********** **** **** significantly ********* ** *** lighting ********** ** *** scene (*******, ********, *****, etc.).
  • ********* **** ** ****** faces - ******* *** determining **** ******* *** in * ***** *** whether ***** ******* *** a **** (******* ** a ****, * ***, a ***, * ******* ball, ***.) *** ** very *********** ***** **** video ************ ******* (*.*., IP *******, ****) **** limited ********** *****.

* ****-*** **** ******* directly ** *** ****** will ** **** ** detect:

***** *** **** ** a **** ** *** lighting **** ** **** challenging:

Facial ***********

****** *********** ******* *** question "***** **** ** that?", ***** ** ********* information **** ************* * crime ** ******** * suspicious ******.

IPVM Image

****** *********** ****** ****** unless *** **** *** already **** ********.

1:1 ** *:*

****** *********** ****** ******** a ****** ** * specific **** (*:*) ** identifies * ****** **** many ***** (*:*).

*:* ******** ** ****** Verification. * ****** ********* photo ** ********* ** the **** *********** ******, and ** ********** ** the **** ******** ******* that *****. *:* ******** is ********** ****:

IPVM Image

*:* ******** ** ****** Identification. **** ****** *** presented ** *** **** recognition ******, *** ** determines ** *** **** matches *** ** ***** photos. *:* ******** ** relatively ****, **** ****** than *:*:

IPVM Image

***** ************ **** *********** applications *** ********* *:* for **************.

Cooperative ** ************* **** ***********

*********** **** *********** ** relatively ****. **** ** when * ****** ***** and ******** ***** ******** at *** ****** (*.*., access *******, ******** ***********,******):

IPVM Image

************* **** *********** ** relatively **** ******* *** person *** ** ********* to *** ****** ** there ** ***** *** know *** ** ****** to ***** **:

********, ***** *** ****** is ********* *** ******** conditions *** ******** *** challenge, ********* ** ************* lower ********.

Logistical ****** ************* **** ***********

******* **** ****** *** unaware **** **** *********** systems ***** *** ************ cameras *** ********* ********* high ***** *** ****** to ***** **** *****, uncooperative **** *********** ** a *********** ********** ** real-world ************.

**** ***** **** ********* good **** *********** ******** with **** ************ ******* is **** ***********.

Resolution / ***** ************

**** *********** ******** ************* higher ********** ** ****** per ****/***** **** ******* person ********* ** **** detection. **** ************* ******* requirements ** "****** ******* the ****" ** "****** per ****". ***** *** specifications ****, **** ************* require ********* ******-***-**** (~*** ******-***-*****).

*** ******* ****-***** ************ cameras, **** ********* ***** short ********* *** ****** accurate ****** ***********:

IPVM Image

Mounting ****** / ****** ********

**** ************ ******* *** installed ** ******** ** above **' *** ****** so **** *** *** of *****. **** ***** that **** ******* *** aimed ***** ** ** the *** ** * person's ****, ***** ***** face *********** *********** *** a ****** ** *********:

IPVM Image

**** *** **** ** capturing ****-**** ******, **** facial *********** ******* **** specific ******** ************ **** minimum ******* ** ******* angle ** **** *******.

Angle ** ********* / ****** ********

******* **** ************ ******* are ********* ** ** next ** *****, ** they *** ***** ** much **** ** * room ** ********, **** do *** ********* ******* a ****** **** ** faces, ***** ********* *** side ** * ******'* head:

IPVM Image

***** **** ****** *********** systems **** **** ** match ***** ** * high ***** ** *********, even **** **** ******* a ***** **** ** so **** ************* ***** accuracy ** ********** ******** to ****-**** ******.

Lighting ******

***** **** ************ ******* are **** ** ******* enough ******* *** ****** detection ** *** *****, facial *********** ******** ************* higher ******* *** ***** clarity.

**** ******* *** ***** low-light ****** ********** **** facial *********** ** *** light **** ***********/**********:

**** ******* ******* ** LEDs ** ********** **** areas *** ******** ******* captured ** *** ******. However,** *********** ****** ****************** *** ******** ******* and ******** **** *********** confidence:

IPVM Image

Avoiding / *********** ****** *********** - *****, ****, **********, ***.

**** ******* ***** *** people ** ******** ** inadvertently ********* *** *********** of ****** *********** *******:

  • **** *** **********
  • *****
  • ******* **** / ****
  • ******

Hats *** **********

********* ** *** ******** (interior ** ********) *** region, **** *** ********** are ******** **** ******* reasonably ******* **********. *******, hats *** ********** ********** impact ****** ********* **********:

IPVM Image

***** **** ****** *********** systems ****** ***** ******* to ******* ****** *********** with ****, ********** *** even *****, ** ****, it ****** ******* *** confidence ***** ** ******* and ********* ********* *** probability ** ******.

*****

***** *** ******** **** common ** ****** ******, not **** ** ******* facilities, ** *** ****** of *********** *********.******* ***** ************* ******** match **********, **** ****** ********* not ********** ** ***:

IPVM Image

*** ******** ** **** recognition ********** **** **** not ******* *** *** decrease ** ***** **** are ******** ***** ******* masks, ***** ***** ******* compound *** *********.

Looking **** ** ***** **** ***********

******* **** ** * simple *** *********** *** to ********* ****** *********** systems. ** ******** ** the ********* ** ******* camera ************ *******, **** people **** ****, ** habit (*.*. ******* ** smartphones):

IPVM Image

***, ************* ********* ** other ****** *** **** to ***** *********, **** known *** ***** **** looking **** ***** ** harder *** ****** ** facial *********** ******* ** identify ****.

IPVM Image

Crowds / ****-******* *********** ******

**** ********* **** **** recognition ** ******** *** (e.g. ********, *******, ********) often ********** ***** ****** of ****** **** ***** issues **** **** **** detection *** **** *********** systems.

** ******** ** *** high ****** ** ***** causing ********, ***** ***** blocked/obstructed *** **** *********** can ***** ******** ********** for *** **** ****** in * ***** **** period.

Liveness *********

****** *********** ** ***** widely ******** ** * solution ** ******** ****** control. *******, **** ******* do *** ***** ******** detection, ***** ***** **** can ** ******* ** believe * ***** ** a ****** ** * real ******:

IPVM Image

******** ********* ** **** commonly ******* *** *** be ********* ** ******* specialized ********. *** ****, see:****** *********** ******* **** Simple ******** ********* ****.

Accuracy *** ****** ***********

********* ******** *** ****** recognition ******* ** **** hard. ******* **** ** market **** **********, **********-******** metrics **** **.*% ********:

IPVM Image

*******, ********** ******* ****** recognition **** ***********:

IPVM Image

***** **** ****** ** beyond *** ***** ** a *** ************, *******, it ** ********* ** understand **** ***** ****** express *** *********** ******** between ***** ********* (*.*., matching *** ***** **** to ******* ****) *** false ********* (******* ** match * **** ** a ****** ** *** watchlist).

IPVM Image****** * *.* ********** standards *** ********** ********** that ***** **** *********** algorithms ** **** ** their *********** *********** ****** ******* (FRVT).

****** **** ******* *** cited ** **** ************* as * ********* ****, however, ** *** ** difficult *** ***-******* ** understand *** ******* ** the *******, ***** *** abstract *** ********.

**** **** ********** ************, the ******* ***** ****** that ****** *********** ******** is ****** ********* ** the ****** ** ** being **** - *.*., how *** ***** *** camera ** ******* *** what ****** *** ******* or ***** **** *** undermine ***********.

Ethical ********

**** *********** *** ******** mainstream ***** ********* ******* of *** ******* ******** of **** ***********, ******* mass **** **********, ******************. ********, **** *********** systems *** ** **** by *********** ** ****** minorities *** ********* ******* for **** ********* *** intrusive ************ (*,*,*).

***** ** ******** ** **** ********* *** *** ********* ** *** ** due ** ****(******** *** ****** ********), companies ************* ******* **** *********** ******* with ****** ***** ********, to ******** *** ***** of **** *********** *** public ******/****** *****.

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

***** ** * **** of ******* **** *********** providers, ****** **************:

  • ******: ****** *****-*****************, **** ************* **** cheaply ***** ***** *** facial *********** *******.
  • *********: **** ****-****** *** controversial ******* ****** **** recognition *** ************, ****** control, *** ****** ************ (**** ****)
  • ********: ***-**-*** ************ ************, recently ***** **** *********** (**** ****)
  • ********: ***** ********** ********* developer, ******** ***** **** recognition (**** ****)
  • *****: ***** ***** ************, offers **** *********** ******* and **** (**** ****)
  • *********: *****'* ******* ************ manufacturer, ****** **** *********** cameras *** **** (**** ****)
  • ******/****++/*********: ***** **** *********** specialists ** *****, **** little ******** ******* *****
  • ***: **** ******** **** for ********** ****** *********** projects *** **** ***********, requires ********** *****.

Comments (9)

*** ***** **, ******* we **** ** ** not, **** **** ********** genie ** *** ** the ****** *** ** unlikely ** **** ** put **** **** **. Face *********** ** ******** more *** **** **********, one **** *** ** look ** ****** ***** and *** ***** ** Facebook ** *** *** indifferent ******** ** *** identification ** ****** ** their *** ******, ** the ****** ** ******.
******** ** ********* ** the ****, ***** ** pushing ******* **** *** "social ******" ******, ***** appears **** *** ******* to **** ******* ** what ** ***** **** public ***** ****** ***********. Yet, ** **** *** seem **** ********** ** this ** ******* ******* are ************ **********. ********* this *** * *** to ** **** ********** themselves, *** *** *********** must ** ********** **** people ******** ***'* ****.
**** ** *** **, certainly ** ** *****, use ** ****** *********** systems ** * ****** environment ** ******* ************ traction. **** *** ***-**** perspective *** ******** ** deploying **** * ****** appear ** ******** *** other *******.

Agree: 2
Disagree: 1
Informative
Unhelpful
Funny

* *** *** ***** of ******** ********* ***** this *******. ***********!

Agree
Disagree
Informative
Unhelpful
Funny

******, ******! ** **** a ****** ** **** -****** *********** ******* **** Simple ******** ********* ****. **'* ** *********** read. *'** ***** **** link ** *** **** of *** ******.

Agree
Disagree
Informative: 1
Unhelpful
Funny

** ****,

*'** **** ******* **** and **** ***** ***** facial *********** *** **** hearing **** ***** ********** can ****** **** ************ reads **** *** ***********. Have *** ****** *****?

Agree
Disagree
Informative
Unhelpful
Funny

** *** ******** ******* to **** ***** *** about ***** ****** ************ about ****** *********** **** masks **. ** **** not ****** **** *** but ** **** ****** with **** ***********.

***, *** *** *** hearing **** ***** *****?

Agree
Disagree
Informative
Unhelpful
Funny

* *** ******, **** customers *** ***** ***** them *** ***** **. In ****** ***** **** their ***** ****** *** a *** **** ** company’s **, **** ** Europe. * *****’* **** or ***** ** *** actual ******** ****** *** wanted ** ***** ****.

Agree
Disagree
Informative
Unhelpful
Funny

** ***** * ******** for ******* ** * camera ************ *** ****** recognition? * ******* *** mounting ****** *** *** other ******* ******* ** this *****.

Agree
Disagree
Informative
Unhelpful
Funny

** ******* ** *** face *********** **********, **** vendors ***** *** **** to ** *******, **** can **** **** ** 45, * ***** **** vendor *** **'* ******/**** points, *** **** ** find *** *** **** suit **** *****, **** are **** ********* ** light ********** ** *** them *** **** * better ****** ** ******* the ************

Agree
Disagree
Informative
Unhelpful
Funny

***** **** ***** ******* IPVM. *** ***: ************ need *** ** * single ***** *** *:*. It ****** ******** **** the **** ** ********** through **** ***** ******** (e.g., **** *** **** or **** *** ****). So ** ** ****** control ***********, * **** may *** * **** or *** ***** *** then *** **** *** can ****** ***** **** lots ** ****** ** the ****** ** ****** the ********.

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

This article is part of IPVM's 7,114 reports and 943 tests and is only available to members. To get a one-time preview of our work, enter your work email to access the full article.

Already a member? Login here | Join now
Loading Related Reports