The Mar­vel of Fa­cial Recog­ni­tion


iD magazine - - Contents -

Are you a su­per recognizer?

There are peo­ple who never for­get a face— and can rec­og­nize oth­ers bet­ter than any com­puter. Could you be one of them?

It’s the big­gest manhunt to take place in London since the 7/ 7 bomb­ings in 2005: When Alice Gross sud­denly dis­ap­pears on Au­gust 28, 2014, the Metropoli­tan Po­lice send 600 of­fi­cers to look for the 14-year- old. They search streets as ca­nine squads probe parks and res­i­den­tial ar­eas and po­lice divers scour the beds of rivers and canals. But it’s all in vain—un­til a new po­lice unit suc­ceeds in crack­ing the case. In a very short amount of time, the 20 su­per-rec­og­niz­ing mem­bers of this spe­cial unit man­age to do what their col­leagues couldn’t: They lead in­ves­ti­ga­tors to the girl’s body and iden­tify her mur­derer, Ar­nis Zal­ka­lns, us­ing only their eyes and minds, and with­out ever leav­ing their desks at New Scot­land Yard…


Su­per rec­og­niz­ers have a unique abil­ity: They prac­ti­cally never for­get a face—and can rec­og­nize it again and again re­gard­less of whether it ap­pears in a blurred pho­to­graph, an in­frared image, or an image in which the face is par­tially ob­scured. It’s a pre­cise de­ter­mi­na­tion process that works like clock­work. To find Alice, su­per rec­og­niz­ers eval­u­ate images from the 300+ surveil­lance cam­eras po­si­tioned in a 2.3-square-mile area around Grand Union Canal, which is where the girl had last been spot­ted.

The team must iden­tify sus­pects from among tens of thou­sands of mostly blurred or in­dis­tinct images. To most peo­ple, the images would just look like a col­lec­tion of pix­els— but not to a su­per recognizer. The rea­son? “Not ev­ery­one sees things in the same way,” ex­plains De­tec­tive Chief In­spec­tor Mick Neville, leader of the Cen­tral Foren­sic Image Team.

I hardly ever for­get a face, and I’m able to rec­og­nize you im­me­di­ately.”

The spe­cial­ists re­trace the route Alice took through London as well as that of the cul­prit seen in a photo, since he was also re­ported miss­ing.


On av­er­age, peo­ple re­mem­ber only one out of five faces. As for a su­per recognizer? “We can’t say for cer­tain that these folks never for­get a face, but they can re­mem­ber peo­ple they had a mere fleet­ing en­counter with 10 years ago,” says Josh P. Davis, a psy­chol­o­gist at the UK’S Uni­ver­sity of Green­wich. These in­di­vid­u­als can even out­per­form a com­puter: Even the most ad­vanced com­puter can’t match up a child­hood photo with an image of the per­son as an adult—a task that is not easy even for su­per rec­og­niz­ers. Nev­er­the­less, in a test con­ducted at Get­tys­burg Col­lege in Penn­syl­va­nia, su­per rec­og­niz­ers did man­age to cor­rectly match up about half of the 56 child­hood photos of celebri­ties shown to them—10 times the av­er­age pop­u­la­tion’s top score. Mean­while, recog­ni­tion soft­ware can fail if the tar­get per­son is gri­mac­ing. De­tec­tive Neville cites the ex­am­ple of the 2011 riots in Cen­tral London, when the soft­ware was used. “Our data­base had 4,000 images and the soft­ware iden­ti­fied one ri­oter. One,” ad­mits Neville. “In the same case, a sin­gle su­per recognizer was able to iden­tify 180 peo­ple.”

The in­ves­ti­ga­tion en­tailed mak­ing the best se­quences from hun­dreds of thou­sands of hours of cap­tured CCTV footage avail­able to the su­per rec­og­niz­ers—as soon as pos­si­ble. “When­ever a se­ri­ous crime such as mur­der oc­curs or there is a threat to pub­lic safety, all mem­bers re­ceive one or more images via smart­phone. But they are not pro­vided with any back­ground in­for­ma­tion about the case so they can’t be un­in­ten­tion­ally steered in a cer­tain di­rec­tion,” says Davis. But just be­cause some­one rec­og­nizes a face in a photo doesn’t nec­es­sar­ily mean this is enough for a sus­pect to be found guilty in court, just as a fin­ger­print at a crime scene

A su­per recognizer can iden­tify 180 times more faces than a com­puter.”

doesn’t prove the guilt of its owner. “But an image is ev­i­dence, and as such, it’s just as ef­fec­tive as fin­ger­prints and DNA,” ex­plains Neville.

It is es­ti­mated that 1 to 2% of the pop­u­la­tion have this spe­cial abil­ity, but most don’t re­al­ize they have it.

“I didn’t know how good I am un­til I took a test,” says su­per recognizer Gary Collins, who’s among the best in his field. There’s no trick to it, and it can­not be learned: “Some­thing in my head just ‘rings’ when­ever I see a fa­mil­iar face,” Collins ex­plains.


You’re ei­ther a su­per recognizer or you’re not. ( Take the test, at right.) To date, these peo­ple are a mys­tery for re­searchers. Just as mys­te­ri­ous: They’ll for­get inan­i­mate ob­jects just as fast as an av­er­age per­son does in tests. Fur­ther­more, some of them have trou­ble re­mem­ber­ing names— they only re­call where they saw the per­son and how he or she looked. “We still know very lit­tle about how these in­di­vid­u­als are able to more

ef­fec­tively uti­lize the brain cen­ters re­spon­si­ble for recog­ni­tion of faces. Nev­er­the­less, we’ve found that their abil­ity seems to be in­her­ited,” says psy­chol­o­gist Josh Davis.

The ba­sic features of a hu­man face can be ar­ranged in about 5 sex­til­lion pos­si­ble com­bi­na­tions, which means in the­ory there are 700 bil­lion times more dis­cernible faces than there are peo­ple on the Earth. This means that every face is like a unique code and su­per rec­og­niz­ers can de­ci­pher it bet­ter than the gen­eral pop­u­la­tion, although we all read faces the same way: “Brain scans have shown that the brain first per­ceives the ba­sic shape of a face. Then it es­tab­lishes the rel­a­tive spac­ing be­tween eyes, nose, etc. Only at the end does it ‘look’ at the face as a whole. Su­per rec­og­niz­ers are bet­ter at this,” says Davis. This abil­ity is in­dis­pens­able to London’s po­lice, and con­vic­tion rates have risen. Above all, the spe­cial­ists save in­ves­ti­ga­tors time—time to find crim­i­nals be­fore they can strike again.

I didn’t know any­thing about it. Then a test showed: I’m a su­per recognizer.”

THE QUICK­EST SENSE It only takes frac­tions of a sec­ond for the brain to rec­og­nize a face— even though there are an es­ti­mated 5 sex­til­lion dis­tinct com­bi­na­tions of pos­si­ble fa­cial fea­tures.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.