Corsight AI's Aggressive Claims Examined

By Robert Wren Gordon and Sean Patton, Published Nov 01, 2021, 10:04am EDT

Corsight AI aggressively claims that its facial recognition software works under extremely difficult conditions like below with a person in a crowd wearing a mask and with his head titled:

IPVM Image

IPVM spoke with Corsight AI's CEO Robert Watts to better understand the company's claims.

**********

IPVM Image

******** **** * *** ****, Israel-based ****** *********** ******* with ********** ** *** UK *** *** **. Earlier **** ****,******** ** ********* ******, *** ****** ************ Camera ************ *** ******* and *****, ** *** Chief ******* *******.

** ********* ******* ******** AI ********* ********* ******** ********** ******, including ***** **** ** recognize ***** ****** *** masks, *** ******** **** ******* ******** "**** Role **** ****** *********** Tech **** ** ****-******** Travel?,"********* ******** ** *** Jersey-based***********.

**** ******* *** ** Corsight ** ********* **. Watts's ****** ** *** TravelPulse *****, ********* ** a **** **** **. Watts, **. ******, *** Corsight ** ********** ******** *****-************ ***** *** ******* demoed *** ****** *********** software.

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

******** **'* ************* ** IPVM ******:

  • ***** ******* (* ****** or ****) ** ***** datasets/watchlists (*.*., ***,***) **** a ****** ****** ******* graphics ****.
  • **** ***** ** ********* (up ** **°) *** recognition.
  • ** ****** "******* *** ears" *** *********** ** ~90PPF.

*** ******** **** *** demonstration *******:

  • ****** ************ ** ********** were *** ************ ****** but ****** *********-***** *********.
  • ***** ******** ** ******** an ******** ******** ** accuracy *** ** *****, hats, ****** ** *********, lighting; ******** ***** *** disclose ******* ** ******** values. *******, *** ******* said **** ** ****** this *********** **** ********.
  • ******* ***** ********** ****** demonstrations *** * ********** low ** ***** ***** increased ***** ******* ** real-world **********.

****** **/**/****, ******** ********* saying " ********* ****** should ******, ***** ********, be ** ****-******* ** they *** ** (*** example, ******** ** "*******-*****"). Corsight **’* ************* ********* are ************** ** **** and **** *** **** in *** **** ***** by *********." ****'* ******** is **** ***** **** is *** ******** *** surveillance ****** *** ****.

******** **** ********* ****** "Corsight: ** ** *** low *** ********’* ******. In ****, *** *********** threshold *** ******* *** typical *** ***** ** 45, * ********* **** results ** * **** low **** ** ***** positives, **** **** ***** a ** ** **,**** suspects."

Percentage ******** ******

******** ******* *** ********* data, **** *** **** 2 **** ************ ****** scenes *** ***** ********:

IPVM Image

******** ******** ** ***** any ******** ** ****** from ***** "*********** *********" datasets ** ** ****** tell *** "***********" / hard **** ***. *** company **** ****:

*** **** ** *** is ******** ** ******* a ***** *********** ** quality ********* *******, ****, occlusions, *** ***** ********** met ** *** ********'* site. ** ***** ** have *** *** ******* as ***********, **** ******* is ******** ** ******* indicators, **** ** *** distribution ** ***** *** yaw ******, ***** ********* and ***** ******* **** can **** ** ************ the ***. *** *******, a *********** *** ***** have * *********** ***** of ***** ****** *** occluded *****. ** **** stage ** ****** ***** specific *********** ********* ***** sets.

Corsight ** ***: "** ***'* **** ***** ******"

IPVM Image

******** **'* ***** ********* defended *** ************ ** IPVM, ******* **** ***** was "** **** ** be ********" ** *** business:

We ***'* **** ***** ****** ** *** ******. *****'* ** **** ** ** ******** ** *** ************ ** *** ********. I was very, very clear when I took over as Chief Exec, that there are organizations in, I suppose, this market whose marketing engines are far, far in advance of their delivery engines. But that will never, that will never be Corsight. [emphasis added]

Ski *****, ****, *** ******* *****

******** **'* *** ** quoted, ** *** ********* of *** *********** *****, saying **** *** *******'* software *** ******** ****** through *** *****:

** *** ******** ******* with **-****** **** ****. We *** ******** ******* with * ****** ** 70-degree ****,we *** ******** ******* **** * **** ****—*** **** **** * **** ****, *** **** * *** ****...with hat, glasses, all of that[.] [emphasis added]

****** **/**/****: ******** *********, saying "******** ** ********** faces **** **° ***** elevation ******* ** **° (outlined *****). *** ******* contained *** ***** ***********, which *** ******* ** IPVM ** * ******** email."

********** **'* ***'* ******** to ***********, ** ** ****** saying **** "** *** identify ******* **** * face ****—*** **** **** a **** ****, *** with * *** ****...**** hat, *******,all ** ****," with the addition of "all of that" implying that Corsight AI is capable of identifying someone wearing a ski mask, hat, and glasses simultaneously.

*******, ******** ** ****** back **** *****, "**********" agreeing **** * ****** wearing *** ***** ***** would ****** ** * significant ********** ******** ***/** missed ***************. ***********, ******** AI's ********** **** ******** his *******'* ******* ** identify ****** ******* *** masks ** **** ** glasses ******* ***** ***********:

***** *** *'* ********** just **** ***,*** ****** live, ** ****-****, ** my ******. **, ***** are ***,*** ****** ** my ********* ***** *** that ** *** **********, I *** *** * hat **, * *** put ******* **, *** system, ** * ***** my **** **********, ** course, *** *'* ******** way *** **** ** the ****, *'* *** going ** **** ** do ***********. *** ** I **** *** *** and *** ******* ***, it *** ** **.

** ********* ******** **'* demo, **** ******* **** Corsight ** ********* ********** scores ** *** ******* and *** ******* ** images ** *********** ******* ski *****, ** **** with *** ***** ********* ** ******* **** report ** ********, ********** * ********** score *** *** ***** of *** ****** *** of **** **%:

IPVM Image

IPVM Image

************, ** ** ***** noting **** ******** ** considers *** **** ********* in *** ***** ** the ***** ** ** a '*** ****,' **** Mr. ***** *********:

[****] ** **** ** would **** ** ****** a *** ****, **** is **** * ***** wear **** * ** skiing.

** *** ****** ******, this ******* ** **** commonly ***** ** * balaclava, **** ******, **, more ************, * ******* balaclava.

Matches ** ***** ******

******** ** **** ******** its ******* ** ******** matches *** ******** ****** based ** *******-*** ******. In *** *********** *****, Corisght **'* *** ** quoted ********:

** *** ******** ******* who ** *** *** 50 **** * ********** of **** ** *** 17.

**** ****** ******** ** marketing ******, ******** ** a ******* ******** ** IPVM:

******** ** ******** ****** dated **** **-**+ ***** (while *** *********** *****’* go **** ** *****).

**. ***** ***** ** these ****** ** *** company's **** **** **** while ******** ****** *****, *** *******'* **** of ******** ***********:

* **** ** *** EVP ** ******** ***********, he *** ****** ** himself ** **. *** he's *** ** *** 50s, **** ***/***** ***. And **** **** ******* against ***. * ****, it ***** ** **, I'm ***** ** *** 45% ** ***** ** same ****? **, **'* a ****** *********, *** it's ***** ****** * match, **'* ***** ****** a ****** *****.

IPVM Image

*******, **** ******** ** demoed ******* ******* ***** photos, *** ******* *** unable ** ******* *** example **** **. ***** and ******* **** * glamor **** ** ******* actress ***** ****** **** she *** **** **–**, showing **** ******** ** software *** **** ** match **** *******-*** ***** to ************ ****** ** the **-****-*** ***********.

*** ********** ***** ** the ***** ****** **** match *** **** ** the *******. ***** ******** describes **** ** "****** a ****** *****", **** a ********** *** ********** would ********* ******** *** number ** ***** *******, especially ** ******.

********, *** ****** *** paparazzi ***** *** *** clearer **** ************-******* ******, as ****** ** ******** AI ** **** ** a ******* ********:

***** **** ******** ***** frames ******* *** ** significantly ***** ******* *** resolution, *** ******* *********, difficult ****** *** **********.

Data "******* ****** *.* ** * ******"

******** **'* *** **** claimed **** *** *******'* software ******* ******** **** within ****** ** * second ** *** *********** interview:

**, ** **** ** my ***** ** **** image ** **** ** a ************ ********—********* ****** the ******** ********—***, ***’** looked **, ***’** ****, butyour **** ** ******* ****** *.* ** * ******. I can guarantee that to you. [emphasis added]

**** ****** **** ******* this *****, ******** *** that ** *** **** of * ***** *****, data ***** *** ** deleted ** ***, * fact ************ ** ******** AI:

***'** ********** *****. ** a ***** ***** ******, then, ** ******, **** image **** ** ******** against **** *****.

******** ** **** **** the **** ***** **** on *** ***** ********, or *** "***** ** the ****," ** ****** falsely ********** *******, ****** that ******* *** ** performed "********* *******, ********* simply." *******, ***** ******* and ***** *********** **** facial *********** ******* *** common (*.*,*,*).

Corsight ** ******** *****-** ********* *****

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

IPVM Image

******* **** *** ***** reveals *** **** "*****-**: detected" *** "*****-**: ********" immediately ***** *** *********** readings ** *** *** and *****, ************.

IPVM Image

******** ******** ** *** not ******* *** **** featuring **** ********** ***** from *** ******** ****, the ******* ********* *** graphic, ******** **** *********** published ** ** *** lead ***** *** *** Corsight ** ********* ******* consulting *** ******* *******:

[*]* ** ********* ** flag **** ******** ** did *** ****** *** does *** *** *** rights ** **** ***** ** *** ******** as * ******* ***** by *** **********.Let ** ********* **** ******** **’* ********** ****** *** **** *** ********/****** *****-**. [emphasis added]

**** *** ******** ****** Corsight **'* **** **** IPVM:

*** ****** **** ********* used **** ******** ***** from *********, ***'* **** where, *** ******* ** our ******. ** **** representation ** *** ***********, we **** **** ***** is *** **** *** system.We **** *** ****** *** **** *** ******. That is not from our system, please remove it. [emphasis added]

***** *********** *** ******* its **** ***** ** a ******* ***** ** US ******* ******** *********, Corsight **'* ** **** attempted ** ******* ****** *******'* ******** ************* ** ******* *** misleading *****-** ********* *****:

*** *** ***** ** the ****** *****, * don't **** *** **** you **** *** ***** of ******** ** ****** media *****, *** **** won't ****** ****** ******* the ***** ** *** article ** *******.

UPDATE: *** ********* */ *********** ******** **** *******

******** ** ******* *** LinkedIn **** ******* *** TravelPulse ********* **** *** Watts ********* *** *********** of **** ****, **** the **** *** *********** to ** ***** ****:

IPVM Image

Comments (6)

* *** *** *'** been ** ***** **** Corsight *** ******* ******* their ********. **** **** been **** **** ** earth *** ********* **** their ********* ** **. They **** *** ******** their ****** ** ******* it's *******. **'** ***** testing *** ***'* ***** to *** ** ******** against ***** ******** *********, but ** ***, **'* performing ************* ****. *** now, *'* ******* ** give **** *** ******* of *** ***** ** the ******** *********.

Agree: 1
Disagree
Informative: 4
Unhelpful
Funny

UPDATE: *** ********* */ *********** ******** **** *******

******** ** ******* *** LinkedIn **** ******* *** TravelPulse ********* **** *** Watts ********* *** *********** of **** ****, **** the **** *** *********** to ** ***** ****:

IPVM Image

Agree
Disagree
Informative: 3
Unhelpful
Funny

**** ** ***** ** Video *********, ** ** integrator, ** **** ** manage *** *********** ************ of *** ***-**** ***** with *** *********** *********** promised ** *** ************.

Agree: 2
Disagree
Informative: 1
Unhelpful
Funny

There's ** **** ** ** ******** ** *** ************ ** *** ********.

**** ** ** ******** this ** ** ****, saying ***** ***** *** loud ** * *******, imo.

******* ** ******** ********** in *** ******** **** skeptics, **** **** ** statement ***** ** **** the ******** ******.

Agree: 1
Disagree
Informative
Unhelpful
Funny

**** ****** ***** ********* and ***** ************ **** our ****, ** ******* is **** * *** of ***** ****** *** not **** ******* ***, e.g. *** ******* ** average ****** ***** ***** to ***** ******, ****** **** ** *** interview:

** **'* **** *** surveillance ********, ****** *********** software ***** ***********’ ********** as **** **** ******* ports ** ********, ******* them ******* * ******** to **** ****** *** appears ** * ********** watchlist—missing *******, ****** ***********, terrorists, ***.

***, *** *** ******* person *** ***’* ***** to *** ******, “**’** not ** *** ********.” Watts *********. “**, ** soon ** ** ***** or **** ***** ** seen ** * ************ database—certainly ****** *** ******** software—yes, ***’** ****** **, you’re ****, *** **** data ** ******* ****** 0.4 ** * ******. I *** ********* **** to ***.”

*** ** **** ************ acknowledged ** **:

** * ***** ***** occurs, ****, ** ******, that ***** **** ** retained ******* **** *****.

**** ** **** **** worry *****. **** **** are **** ******* **** the ***** *** **** are ******* ****** **. Or * ******* **** worse, ******* **** *** to *** ********* (*** you *** *** * terrorist ** ****** **********, etc.) *** ***** ******* are ******* * ***-**** search *** ***.

** ** *****, * don't ***** **** ******** is ***** **** *** facial *********** ** ***** than ***** ****** *********** suppliers *** ***** * defense ** "**** **** is ******* ****** *.* of * ******. * can ********* **** ** you" ** ** **** naive *** **** *** address *** ******* ***** here.

Agree: 2
Disagree
Informative
Unhelpful
Funny

******** *** *********, ****** below:

***** *** *** ********* a **** ** *** article.

** ***** **** ** express *** ************** ** the **** ** *** piece, ***** ** *** view **** *** ******* a **** ********** ** Corsight. ************* ** *** ***** within *** **** ***** article,****** ** ******** ********* stressed**** ******** *** *** supply ** **** *** own *** ****** ** the *****, *** **** is ***** ********** ****** the *****.

*** ******** * ******** from **** ******, ***** Privacy ******* ** ******** AI, ***** ********* **** of *** ******** ****** by** *********.

****** ***** *** ****** within ************** ******** **** *** inaccurate. ******** ***** ********** it ** **** ***** rectify ***** ****** ** soon ** ********.

*** ** **** ** you **** *** *********.

*******,

*****

****: “*** ******** **** the ************* *******:

****** ************ ** ********** were *** ************ ****** but ****** *********-***** *********.”

********: **** **** **** not **** *****. ********* images ****** ******, ***** possible, ** ** ****-******* as **** *** ** (for *******, ******** ** "mugshot-style"). ******** **’* ************* databases *** ************** ** this *** **** *** used ** *** **** world ** *********.

****: “***** ******** ** reported ** ******** ******** in ******** *** ** masks, ****, ****** ** incidence, ********; ******** ***** not ******** ******* ** specific ******. *******, *** company **** **** ** shares **** *********** **** partners.”

******** ** ****** *** data ********* ***** *********** on ***** ******* ** images ******** ** ******.

****: “******* ***** ********** during ************** *** * relatively *** ** ***** risks ********* ***** ******* in ****-***** **********.”

********: ** ** *** low *** ********’* ******. In ****, *** *********** threshold *** ******* *** typical *** ***** ** 45, * ********* **** results ** * **** low **** ** ***** positives, **** **** ***** a ** ** **,**** suspects.

****: “******** **'* *** is ******, ** *** beginning ** *** *********** story, ****** **** *** company's ******** *** ******** people ******* *** *****:

"** *** ******** ******* with **-****** **** ****. We *** ******** ******* with * ****** ** 70-degree ****, ** *** identify ******* **** * face ****—*** **** **** a **** ****, *** with * *** ****...**** hat, *******, *** ** that[.] [******** *****]”

******** ** ********** ***** with **° ***** ********* instead ** **° (******** above). *** ******* ********* the ***** ***********, ***** was ******* ** **** in * ******** *****.

****: “*** ********** ***** on *** ***** ****** demo ***** *** **** in *** *******. ***** Corsight ********* **** ** "giving * ****** *****", such * ********** *** percentage ***** ********* ******** the ****** ** ***** matches, ********** ** ******.”

********: **** *** ************ on * ********, ***-**** scenario ***** ******** **** a **** ** **** old ***** *** ****** it ***** ******** *** person ****** *** ********, where ***** ********* **** evaluate *** *** *** candidates - *** **** the *** * ***** they ** ** "****" - ********* *** **** threshold *** ******** *** to * *** **** purpose ** ****** *** top ** - ****** were *** *********.

** ** ***** ****** that ********** ** *** range ** **+ ** fact ******* **** ********* in **** ** *** specific *******. **** ********* exceeds *** ********** ****** by **** ****, ***** resulted ** * *** of */*,***,***.

*'** ***** **** ** these ********* ** *** body ** *** ******, also ****** *** ******** to ****, *.*. ** Corsight's ******** **** ********* images ****** ** **** quality, ****'* ******** ** that ***** **** ** not ******** *** ************ photos *** ****.

** *** ********'* ********* about *** ***** ********* image **** "****** ** ******** ********* stressed**** ******** *** *** supply ** **** *** own *** ****** ** the *****, *** **** is ***** ********** ****** the *****." ** ***** that ** *** ****** but ** **** ********** noted **** **** **** that ***** ** ********* for ***** **** ***** I ****** **** ******** even ******* **** ****** in *** ******** ****, who ******* **.

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

This article is part of IPVM's 7,330 reports and 971 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