South Wales Police facial recognition software boasts 8% success rate

The South Wales Police Force is defending the decision to trial NEC facial recognition software during last year’s Champions League Final as it is revealed only 8% of the identifications proved to be accurate.

The project was announced last year as the South Wales Police Force outlined plans to use NEC’s NeoFace Watch facial recognition software platform to increase the efficiency of police work during a weekend which saw 170,000 football fans in Cardiff. The promise of the technology was to identify persons of interest on pre-determined watchlists in real-time, with data being collected using CCTV cameras mounted on a number of police vehicles.

While the software was used at a number of different events in the capital, the Champions League Final got the most attention. As you can see from the table below, none of the results are particularly flattering for the South Wales Police or NEC, the firm which provided the technology to power the failed project, with the overall success rate just 9%

Event True Positive Identifications False Positive Identifications Success (%)
Champions League Final 173 2,297 8%
Elvis Festival 10 7 58%
Operation Fulcrum 5 10 33%
Anthony Joshua Fight 5 46 9%
Wales vs Australia Rugby 6 42 12.5%
Wales vs Georgia Rugby 1 2 33%
Wales vs New Zealand Rugby 3 9 25%
Wales vs South Africa Rugby 5 18 21%
Kasabian Concert 4 3 57%

While it is completely understandable there will be flaws in trials and POCs, this demonstration of outright failure makes you question whether this software should have been released from the lab in the first place.

Governments around the world are seemingly becoming less trusting of their own citizens on a daily basis, meaning more intrusive and secretive means of monitoring individuals are likely to become more common. Politicians and spooks around the world must have been watching these trials with some interest, and such a catastrophic failure of the technology is a very worrying sign.

Technology companies like NEC will be under pressure to produce platforms such as NeoFace Watch as quickly as possible as governments continuously look to step up activities. This pressure might result in platforms being launched too early, before enough stress tests have been run. This would certainly seem to be the explanation here, though South Wales Police (and NEC presumably) has blamed poor images supplied by UEFA, Interpol and other partner agencies. That said, the following extract from the NEC website seems to contradict this statement:

“NEC NeoFace technology’s strength lies in its tolerance of poor quality. Highly compressed surveillance videos and images, previously considered of little to no value, are now usable evidence and leading to higher rates of positive identification. With its proven ability to match low resolution facial images, including images with resolutions down to just 24 pixels between the eyes, NEC’s NeoFace technology outperforms all other face recognition systems in matching accuracy.”

It does seem the world has accepted its fate when it comes to Big Brother and eyes in the skies, but companies like NEC need to step up the game. Such technologies are likely to play a role in the trials and potential convictions of individuals in the future, therefore accuracy needs to be as high as possible.

MegaFon and NEC claim network automation victory

MegaFon and NEC have announced the completion of a live field-test to incorporate AI technology into the Russian network to improve the efficiency of planning and maintenance of transport network resources.

The tests took place in MegaFon Ural’s network from October to November, with the NEC AI algorithms analysing 150 radio links which were considered the most critical to the network. The test itself was considered a successful with the telco seeing benefits for both the demand forecast and predictive maintenance objectives.

“Our partnership with NEC aligns with our goals of efficiently improving the planning and maintenance of networks, which is becoming increasingly complex,” said Anton Sherbakov, Technical Director for MegaFon Ural. “Through analytics of big volumes of data with NEC’s AI technologies, ‘NEC the WISE’, we have verified that significant improvements can be achieved for the planning and operation of transport networks, resulting in more effective use of resources and providing the highest quality services to millions of subscribers.”

“We are very pleased to see our services streamline and optimize MegaFon’s network without the need to expend additional resources,” said Hiroshi Kawada, MD of NEC Neva

Communications Systems. “We aim to enhance our partnership with MegaFon and to continue delivering the latest innovations to the entire operator network.”

Looking at the demand forecast side of the test, the pair claim the AI component was 97% accurate when it came to traffic prediction against actual traffic. Predictive maintenance is a bit more difficult to justify on the spreadsheets, if the algorithm is accurate the negative instance never actually happens, but both seem happy with the results.

‘NEC the WISE’ is a product line up which was announced back in 2016 but now with the current euphoria surrounding AI, it seems to be making some useful headlines for NEC. While those who are stoking the 5G fire seem to be focusing on generating new revenue channels for the telcos, this product line has been billed for efficiency. This portfolio has been in the works for the last couple of years, but is now being pitched to clients, but NEC has also stated it plans to deliver a fully automated network operation solution by 2020.

Whether a fully autonomous network is achievable by 2020 remains to be seen, but the normalization of more optimized networks might be more accurate.

Biometrics could be a threat to the smartphone – NEC

Our lives are dominated by a tiny digital interface, but biometrics could offer future generations a more natural way to interact with the digital world.

Before hysteria develops it should be worth noting this piece is not predicting the death of the smartphone, but Shinya Kukita, Chief Engineer of the Global Business Unit at NEC pointed out it is not a very natural way for humans to communicate. You’re hunched over, prodding a tiny screen and sometimes squinting to pick out the details. We’ve gone from a very expressive, vocal being to one which is dominated by silent communication and limits. Of course there is some way to go, but the steps forward being made by biometric authentication mean the world could become a lot bigger.

“With so many displays on the street, based on location or biometric details public screens could become a personal display for that particular moment,” said Kukita. “Any screen could become personalised.”

The accuracy being shown by biometric authentication technologies is increasing quickly. At the NEC stand we saw a quick demonstration of its facial recognition tech and the software couldn’t be fooled by a picture of the individual or a video, and the software was even able to detect whether the individual was very an incredibly detailed mask. There are becoming fewer ways to ‘spoof’ the software according to NEC.

With the public display scenario, once there are sensors everywhere any individual could work up to a screen on the street and it would recognise that person. Maybe a two-part authentication would be needed to make 100% sure (choose from facial, voice, fingerprint, palm, iris, ear canal), but because all your data is stored in the cloud, this public display becomes your personal screen for the moment. It would not necessarily be the death of a smartphone, but it could decrease our reliance on it.

While it is feasible, there are more practical applications for the technology right now. Back when the Champions League took place in Cardiff in last year, the local police used NEC’s facial recognition software, known as Bio-IDium, to identify individuals who had been placed on a watch list. Perhaps these individuals were banned from the stadium, or they had been involved in a recent incident, but Kukita highlighted several arrests were made once they had been identified and the nearest officers informed.

This is a very specific usecase, and in many European countries only the police force have been enabled to use such technology, but there are wider applications. Display advertising on the street or screens in bus stops could be personalised to the individual who is standing in front of it. Using powerful AI technology, this advertising could go further, perhaps logging into your schedule and realising you need to be on the other side of town. All of a sudden the screen has ordered a taxi for you.

One of the big problems with these technologies is privacy. “There is a fine line to walk between privacy and security,” Kukita commented, and the rollout will have to be measured and conscious. The risky aspect of the facial recognition software is the fact it doesn’t have to be ‘opt-in’. With a finger print scanner you essentially give permission by offering your finger to the machine, but facial recognition can operate in the background without the knowledge of the user. It is a delicate balance to strike, hence why only police are legally allowed to use it right now. One major incident and public opinion could be influenced, setting AI progress back years.

There is a lot which needs to change in terms of regulation, a huge amount of progress to be made deploying sensors and a monstrous amount of public consultation which will need to take place before the world we have described is a reality, But, it isn’t really that far away.