SUPER RECOGNISERS

Still outperforming AI algorithms, the Met’s super recogniser unit now aims to tackle cross-racial bias

Adam Bannister

Editor, IFSEC Global

Author Bio ▼

Adam Bannister is editor of IFSEC Global. A former managing editor at Dynamis Online Media Group, he has been at the helm of the UK's leading fire and security publication since 2014.
April 23, 2018

Sign up to free email newsletters

Download

The Intruder Alarm Report 2020

Prosopagnosia – or face blindness – is a neurological disorder characterised by an inability to recognise faces.

Studies have suggested that up to 2% of the UK population – roughly 1.5 million people – might have developmental (as opposed to acquired, through brain injury usually) prosopagnosia.

A pioneering 2009 study led by Harvard psychologist Richard Russell found that a similar proportion of the population “are about as good” at face recognition and perception “as many developmental prosopagnosics are bad.”

The oft-repeated phrase “I never forget a face” suddenly had greater resonance.

Russell eventually coined the term ‘super-recognisers’ to describe this minority with exceptional facial recognition skills.

Coincidentally, at around the same time, DCI Mick Neville of the London Metropolitan Police noticed that some of his officers could pick out a suspect from the Met’s ‘Caught on Camera’ notices based solely on a single photograph they’d seen weeks or months previously.

Speaking at the NSI Summit 2018, which I attended recently, Neville recalled how these newly discovered skills were soon put to the test.

London riots

Following the riots that broke out across London in August 2011, Met officers identified 609 suspects from tens of thousands of hours of CCTV footage. One officer, PC Gary Collins, made 180 identifications alone – including a man who had covered his mouth and nose with a bandana and pulled a beanie hat over his forehead. Collins hadn’t seen the man for several years.

Emboldened by this success, Neville helped New Scotland Yard assemble an elite unit of super recognisers, whose ranks now number 140.

Super recognisers can recall up to 95% of faces they’ve previously seen, compared to 20% achieved by the average person

The UK has wasted billions of pounds on CCTV cameras whose footage goes to waste for want of the right supporting skills, Neville told attendees at his NSI Summit talk, which took place at the Vox in Birmingham. Super recognisers, who can recall up to 95% of faces they’ve previously seen compared to 20% achieved by the average person, give the taxpayer better value from their CCTV investment and help police put career criminals behind bars.

They can improve detection rates of repeat offenders, pick out people engaged in suspicious behaviour, identify touts and thieves at major events and spot fans banned from football stadia.

The unit still gets impressive results when footage is grainy, lighting is poor or camera angles are awkward. Decades after the Hillsborough disaster, 16 victims were identified in footage of the stadium for the first time.

Super recognisers played a pivotal role in solving the murder of teenager Alice Gross in 2014.

They viewed thousands of hours of less-than-optimal CCTV footage and within days identified the schoolgirl and suspect Arnis Zalkalns. The timeline that took shape following their painstaking work eventually led to the discovery of the schoolgirl’s body in the River Brent.

Austin Caballero stole more than £100,000 worth of luxury goods during a crime spree lasting two and a half years before the Met’s specialist unit picked out the thief from dozens of clips

Neville also cited the example of a man convicted of 43 thefts. Austin Caballero stole more than £100,000 worth of luxury goods during a crime spree lasting two and a half years before the Met’s specialist unit picked out the thief from dozens of clips.

After his conviction DS Eliot Porritt of the super recogniser unit said: “My unit scans thousands of images in order to link offenders to their crimes, resulting in a greatly reduced cost to the public of just one court process as opposed to the dozens of separate cases that could have resulted had he not been recognised and linked by my team.”

Porritt himself recalls how he caught a glimpse of his remarkable abilities in childhood. “As a boy, I watched The Terminator and Aliens with my father,” he told the New statesman. “I now remember him being amazed when I noticed that an actor – Bill Paxton – was in both films, even though he looked different in each role. But I didn’t think too much of it at the time. I assumed everybody saw what I saw.”

Cologne 2015

German police drafted in the unit, which remains the only one of its kind in the world, to help identify culprits implicated in the sexual assaults on hundreds of women in Cologne on New Years Eve 2015.

The unit has also received visits or requests for help from enforcement agencies in India, Australia and the US, among other countries.

At the NSI Summit, Neville sought to dispel any notion that facial recognition software was ready to put the super recognisers out of work. Facial recognition software identified only one culprit in the 2011 riots, compared to 609 by the super recognisers., he pointed out

Although that was now seven years ago, facial recognition still has a huge gulf to bridge, he insisted. Humans have a particular advantage in assessing people from side-on views and can even identify people from the back of their head alone.

Between 2013 and 2017, Face Recognition Vendor Tests (FRVT) run by the National Institute of Standards and Technology (NIST) found that the False Non-Match Rate (FNMR) improved from 0.068 to 0.025 (although it should be noted that there were changes in testing methodologies over that period).

Neville, who has previously said it could take 10 to 20 years before software is advanced enough to compare with its human counterparts, said technology is all well and good, but there’s not enough focus on people and processes in security.

“Algorithms will get better […] but people change appearance and we as humans are primed to see through those changes.” Josh Davis, psychology lecturer, University of Greenwich

Far from being replaced by tech, he said, the super recogniserrs will only become more effective with the help of analytics tech like pattern and logo recognition software.

Josh Davis, a University of Greenwich psychology lecturer, agrees. “Algorithms will get better and we will be able to build 3D representations of faces,” he told the New Statesman. “But people change appearance and we as humans are primed to see through those changes.”

Super recognisers make up about 1% of the population – you could be one yourself without realising it. Neville invited NSI Summit attendees to have a go themselves, asking them to match various suspects to the right faces in a number of line-ups. (This writer will not be abandoning his day job for a job in the Met).

Neville is CEO at Super Recognisers International, which deploys covert teams on the ground, helps to find missing persons, helps retailers identify prolific shoplifters, conducts post-event analysis (for which the London riots set the template) and trains law enforcement, intelligence agencies, the military and private sector to identify and develop super recognisers from within their own ranks.

The Met augments its elite recognisers’ innate abilities in facial recognition with training in behavioural analysis, facial mapping and gait analysis.

Neville said they were currently trying to improve detection rates in footage with a lot of motion. He also revealed a recruitment drive among ethnic minorities to compensate for ‘cross-racial bias’ – studies have shown that people are generally better at recognising faces of their own ethnicity.

It’s fair to say the police, courts and general public are suitably impressed by the super recognisers. More surprising was the revelation by Nevill that even one of the criminals snared by the unit confessed his grudging admiration. He relayed his congratulations from prison.

The Seagate Surveillance Storage Survey Report 2018

IFSEC Global Exclusive: The State of Surveillance Storage

Download this eBook to begin navigating the fast-changing surveillance and security landscape, from the growing quantity of data to new innovations like Artificial Intelligence and machine learning.

Identify the challenges, and discover exactly what storage solutions need to offer to meet the evolving needs of security industry professionals, installers and integrators.

Related Topics

1
Leave a Reply

avatar
1 Comment threads
0 Thread replies
0 Followers
 
Most reacted comment
Hottest comment thread
1 Comment authors
Michael Neville Recent comment authors
  Subscribe  
newest oldest most voted
Notify of
Michael Neville
Guest
Michael Neville

I have now retired from the police and am the CEO for Super Recognisers International. We can make your venue safer and prevent and detect crime. Email me at [email protected] for information on our services or see the website at superrecognisersinternational.com On 23rd May in London the Association of Super Recognisers, a not-for-profit company to regulate SRs and introduce standards is being launched.