Improvements in Facial Recognition Systems - BUT Still Humans are Needed

A report by the National Physical Laboratory in the UK has found that there has been a “substantial improvement” in the accuracy of automated facial recognition systems and an elimination of bias against women and ethnic minorities.  The NPL noted that for watchlists of 1,000 or 10,000 people, “true positive identification rate” of live facial recognition was 89%. Previously, for a watchlist of 2,000 to 4,000 people, it was 72%.

BUT according to the Surveillance Camera Commissioner’s Code of Practice, “human intervention” is required before arrests are made as a result of a computerised identification. AND the Commissioner has further stated that: “Senior (police) officers should ask themselves how they select, train, instruct, deploy, manage and analyse the performance of the human decision maker…”

Who is better to perform this role than a HUMAN Super Recogniser? According to the US agency NIST, combining Super Recognisers and computerised facial recognition produces “near perfect results”. Unfortunately, too often, senior commanders allow any officer to verify the identification. This is putting the use of facial recognition at further risk and, more importantly, can result in innocence people being detained, whilst the guilty escape justice.  The Association can help you to find your Super Recognisers.

Report on this link:

https://science.police.uk/research/resources/operational-testing-of-facial-recognition-technology/