The world’s most accurate facial recognition software may have already snapped you.
It can scan a photo taken 30 years ago and match it to someone captured on CCTV yesterday.
It’s not fooled by weight gain, ageing and facial hair and can even recreate a face from a skeleton with “remarkable accuracy”.
Australia, the sci-fi police revolution has arrived. And it’s starting in the Northern Territory.
NT Police Minister Peter Chandler said the software, named NeoFace Reveal, was trialed across the Territory earlier this year and identified 300 people.
“It’s extremely accurate,” Chandler said.
“We put photos of someone in their 20s through the system who is now 50 or 60 and the damn thing gets it every time. It's amazing."
More than 100,000 mugshots have been uploaded into the database and he said there was potential to cross reference with health data.
“During the trial, we had a gentleman come into hospital who was completely unconscious,” Chandler said.
“No one had any clue who he was. We snapped him and put it through the software, and got a match.
“If you imagine someone with Alzheimer’s was picked up, it’s my understanding we could work with the health department to use their records to identify them.
"Also in the territory, you wouldn't believe the number of people who gives police a bogus name. This means we'll know if the face matches the name."
The software, which was used to accurately identify one of the Boston bombers, relies on a complex pattern detection tool that measures things like the depth of an eye socket or the length of a chin and can be used on low-res images.
In the U.S. the same software is used by the Arizona Department of Transportation to identify people based on drivers’ licences but Chandler said that raised privacy issues.
“When people give their personal data to the government because they have no choice -- they need a licence -- it’s our responsibility to keep that private,” Chandler said.
“We’re not going to run people through the system who have never committed a crime.”
The software can send real-time alerts when an individual is identified. Picture: NEC
University of Technology Sydney School of Software lecturer Sri Madhisetty said that while privacy concerns remained, facial recognition software was becoming more commonplace.
“If you look at something Facebook, even it is able to identify people in photos using software,” Madhisetty said.
“I think it’s more socially acceptable to search a mugshot database than drivers’ licences because of the expectation of how it is to be used but generally, people accept a driver's licence will be used to identify them.”