Glasgow must copy New York as former beat cop shocked at open criminality in wake of BBC documentary EXCLUSIVE: David Kennedy, general secretary of the Scottish Police Federation, has dismissed suggestions a recent Disclosure episode was unduly harsh on Scotland's biggest city A former police officer who spent years on the beat in Glasgow has said the depiction of the city in a recent BBC documentary was accurate, as he called on city bosses to learn from New York. David Kennedy, who is now general secretary of the Scottish Police Federation, said he counted five arrestable offences in a short walk down Union Street during a trip to the city centre this week. The lack of officers patrolling the streets had left the city a pale shadow of its former self, he suggested. It comes after the Disclosure episode shone a light on drug taking, violence and other anti-social behaviour after film crews spent a week recording there in June, leading to claims the city had become "lawless". The episode has been blasted by Scottish nationalists but Mr Kennedy told the Scottish Express that the scenes filmed were familiar to rank-and-file officers. He said claims that statistics show crime was down in the city didn't "show the full picture," adding: "What shows a true picture is when you look in walking the streets. "I was in Union Street in Glasgow at half past 11 in the morning on Monday, and I saw at least five people who, if I had been in uniform, they would have probably been arrested and given the jail. I was a cop on the beat in Glasgow and I would have absolutely went up, spoke to them, went over and dealt with them." Asked what they were doing, he replied: "Shouting, swearing, obstructing the pavement.
Aug 23, 2026 · via scottishdailyexpress.co.uk
Officers investigating the theft of a shopper’s phone from a store want to track down this man. The victim was targeted in the Wollaton Co-op after she stopped at the till to pay for her items. Having put her phone on the side, the woman left the shop in Lambourne Drive without remembering to pick it back up again. Realising her mistake, she went back into the store, by which time the phone was nowhere to be seen, so she informed shop staff. Having reviewed available CCTV, shopworkers saw that an unknown man had spotted the phone next to the till, pocketed it, and left the store. The theft took place around 7.05pm on 14 July, with the police launching an investigation after being notified soon after. Officers can now release this picture of a man they’d like to speak to as part of their inquiries. PC Kevin Kirk, of Nottinghamshire Police, said: “This incident was sadly an example of how opportunistic thieves can be. “Having spotted a phone left at the till, the person responsible didn’t do the honourable thing and hand it in but instead chose to steal it. “The victim was understandably left very upset by what happened, so we’re determined to try and return their stolen property to them. “On that note, we’d ask that the man pictured does the right thing and makes themselves known to us, as we believe they may be able to assist with our investigation. “Likewise, we’d ask anyone who recognises them to share this information too.” The police can be contacted directly on 101, quoting incident 732 of 14 July 2026, while Crimestoppers can be called anonymously on 0800 555 111.
Aug 22, 2026 · via westbridgfordwire.com
Offering a variety of advertising and sponsorship options for reaching influential specialists from targeted demographic splits.
Cureus provides an equitable, efficient publishing and peer reviewing experience without sacrificing publication times.
Generate broad awareness and deliver relevant, peer-reviewed clinical experiences directly to potential customers.
Dedicated Cranial Radiosurgery: Clinical Experience with New & Innovative SRS Technologies
Sponsored by Zap Surgical Systems
Real-Time Adaptive Motion Management on Helical and Robotic RT Platforms
Sponsored by Accuray, Inc.
Cureus Journal of Medical Science
Sponsored by Zap Surgical Systems
Sponsored by Accuray, Inc.
You can unsubscribe anytime.
By joining Cureus, you agree to our Privacy Policy and Terms of Use.
Aug 22, 2026 · via cureus.com
Mark Zuckerberg’s Meta has never just come out and said that its overarching goal as a company is to boil every last ounce of joy out of the basic experience of being a social animal, but the “boil-joy-and-suck” aspects of their business plans are easy to infer. Even beyond *gestures vaguely, but emphatically, at the last 20 years of online life*, it’s hard not to look at the company’s most recent patent—which would allow the AI-assisted facial accessory the company would really like people to stop calling “pervert glasses” to analyze all of your interactions with friends and loved ones, run them through facial recognition software and other analysis, and then feed you a “highlight reel” of your own life—and not think, “Oh, this was made by people who don’t think too much of the whole human experience, huh?” Per 404 Media, Meta filed the patent for “Smart Cameras Enabled By Assistant Systems” back in February, before it was published last week. In essence, the patent would allow users of Meta’s allegedly smart glasses to turn the whole world into Facebook: Your glasses would record the people around you, tag them, identify when they do things, and then later use that information to present you with a social media-style automatically generated slideshow of things you’d already lived through. (The company’s language repeatedly uses a dinner party as the event our hypothetical memory-deficient voyeur would be ruining.) Besides its obvious and horrifying applications for things like police surveillance and other Flock-esque attempts to shove ever larger portions of the world into monitored cameras—because we’re pretty sure the facial recognition and action-tagging aspects of these things aren’t going to stop at, say, “passing that really good cab sav Michael picked up in Tuscany last fall”—the patent itself is also a potent psychological
Aug 22, 2026 · via avclub.com
At a liquor store, bar, or casino, getting carded is part of the cost of entry. Now the Internet is carding people, too – and the companies doing the carding are cashing in. New laws and platform rules are pushing games, social platforms, AI products, and adult sites to replace birthday boxes (the fields where you enter your date of birth) with identity checks. What was once a compliance tool for sellers in age-restricted industries is becoming infrastructure for the Internet, and a growth market for the companies supplying it. Socure, a Nevada-based identity-verification company founded in 2012, says it now generates more than US$340mil (RM1.39bil) in annual recurring revenue from over 3,000 customers. New ARR (annual recurring revenue) grew 62% year over year, and age verification is among its fastest-growing product lines. The roughly 500-person company has raised approximately US$646mil (RM2.64bil) in private funding. London-based Yoti, founded in 2014 by Robin Tombs, 57, and Noel Hayden, 55, says it has completed more than one billion age checks for roughly half a billion people. Its revenue rose 62% in 2025 and reached roughly US$43mil (RM175.91mil) in the year ended March 2026. And San Francisco identity startup Persona, founded in 2018 by Rick Song, 35, and Charles Yeh, 33, raised US$200mil (RM818.20mil) at a US$2bil (RM8.18bil) valuation in April 2025. The roughly 700-person company has raised about US$418mil (RM1.71bil) to date, and its annual recurring revenue surpassed US$100mil (RM409.10mil) in 2025, according to a Forbes estimate. Age assurance is one use-case within its broader identity-verification business; Persona works with businesses including Roblox, OpenAI, Coursera, and Lime. All three are benefiting from the same shift: Businesses increasingly need to know not just who a user is, but how old they are and which parts of their platform they should be allowed to
Aug 22, 2026 · via thestar.com.my
This post was co-authored with Max Silfverberg (Data Scientist, AI Solutions Lead), Antti Hallavo (Lead AI Software Engineer), and Pontus Huotari (Lead Data Scientist). We work at Alma Media, a Finnish digital services, marketplaces and media company. One of our focus areas is developing AI/ML solutions for real estate listing services, where understanding image content plays an important role. Our solution is to automatically tag photos with room-type and content classes. Our room types include LIVING ROOM, KITCHEN, and BEDROOM. We also tag schematic content like floor plans and site plans. Additionally, we recognize realtor marketing materials, aerial shots, and garden photos. Altogether, there are 23 classes. As Figure 1 shows, this is a classic multi-label classification task; the same space can encompass several room types at once. On the face of it, this sounds simple, but we need to make some tricky decisions. How should you treat a living room photo that shows a bedroom through a doorway? What if the photo only shows 10% living room and the remaining 90% is dining area? The answers depend on the application. If we need to find all photos showing kitchens, we also want to identify living room photos that show a kitchen in the background. However, if the user specifically asks for kitchen photos, we only want to show the ones where the kitchen is in focus. To help decide what to return, classification confidence is important. But depending on how you implement your classifier, you might not have access to that information. Image classifiers can be built in many ways. The modern default approach is to run images through a third-party API which internally uses a vision-language model (VLM) to analyze images and generate tags according to a prompt. Another option is to train image classifiers on top of
Aug 22, 2026 · via towardsdatascience.com
Legal Wrongly Arrested Peppermill Casino Guest Seeks Names of 168 Others Flagged by Facial Recognition Posted on: August 21, 2026, 01:44h. Last updated on: August 21, 2026, 01:50h. A Nevada man who was wrongly arrested after being misidentified by facial recognition technology at Reno’s Peppermill Casino Resort is seeking the identities of 168 other people who were allegedly flagged by the system. Jason Killinger has asked a federal judge to order the City of Reno to produce unredacted arrest reports and declarations of probable cause involving individuals identified by Peppermill’s facial recognition software. The request forms part of Killinger’s lawsuit against Reno and police officer Richard Jager over his September 2023 arrest. Nightmare Ordeal Truck driver Killinger was flagged as a “100% match” to an individual named Michael Ellis who had been banned from the venue months earlier for sleeping on the premises. Despite Killinger’s insistence that he had been misidentified, Jager refused to believe he was who he claimed to be and accused him of using fraudulent identification when he produced a valid Nevada Real ID, Peppermill player’s card, and debit card in his name. Jager chose to trust in the powers of facial recognition software over material facts, such as that Killinger is four inches taller than his doppelganger and has blue eyes rather than hazel. Killinger was jailed for nearly nine hours and spent more than three hours in handcuffs, which left him with bruising and shoulder pain, according to the lawsuit. The lawsuit alleges that even after a fingerprint check at the Washoe County jail confirmed Killinger’s true identity, Jager filed a police report claiming Killinger had presented conflicting identification to Peppermill security. After settling out of court with Peppermill, Killinger sued Jager and later added the City of Reno, alleging the city failed to properly
Aug 22, 2026 · via casino.org
Amnesty International on Wednesday accused Argentina of building techno-authoritarian surveillance infrastructure without adequate legal safeguards, as the government’s newly created AI security unit expands state monitoring powers over citizens’ online communications and physical movements. In its latest report, the rights group stated that Argentina‘s Ministry of Security spent at least USD$1.2 million on surveillance technology between 2024 and 2025, which included social media monitoring tools, facial recognition software, and thermal-camera drones with automatic tracking and ability to transmit images in real time. Amnesty expressed concern for the surveillance of protesters, journalists, activists, human rights defenders, migrants, young people, and marginalized groups. They warn that this can create a “chilling effect,” in which people who believe they are being monitored may avoid speaking, protesting or organizing, even without being directly arrested or punished. This raises concerns about the impact of surveillance on the right to freedom of expression, freedom of peaceful assembly, and freedom of association. In October 2025, Argentina bought a Clearview AI facial recognition license for USD$33,500. Clearview’s database contains more than 70 billion facial images scraped from the internet, and data protection authorities in France, Italy, Greece, and the Netherlands have already fined the company over its data practices. The Ministry, however, maintains that the Federal Police use facial recognition only in connection with active judicial investigations, and no evidence has surfaced that Clearview was used specifically to identify protesters. Argentina’s expansion of AI-driven surveillance has largely occurred through executive and ministerial measures. On July 26, 2024, then-Security Minister Patricia Bullrich signed Resolution 710/2024, creating the Artificial Intelligence Unit Applied to Security (UIAAS) inside the Ministry of Security’s Directorate of Cybercrime and Cyber Affairs. The resolution authorizes the unit to patrol open social media, applications, websites, and the dark web, and permits real-time facial recognition through CCTV footage,
Aug 22, 2026 · via jurist.org
About
Press
Copyright
Contact us
Creators
Advertise
Developers
Terms
Privacy
Policy & Safety
How YouTube works
Test new features
NFL Sunday Ticket
© 2026 Google LLC
Aug 22, 2026 · via youtube.com
Nottinghamshire Police deployed live facial recognition technology in Nottingham city centre for the first time today, Friday 21 August.
The marked police van, equipped with a 360-degree dual-camera system, was stationed in Old Market Square as officers monitored the surrounding area.
The cameras scan the faces of people passing the vehicle and compare them with a police watchlist containing images of individuals suspected of serious offences or wanted for safeguarding reasons.
View this post on Instagram
When the system identifies a potential match, an alert is sent to officers, who must review it and decide whether to approach the person. Police say scans that do not produce a match are deleted within seconds and are not retained.
Signs were displayed around the deployment area informing members of the public that the technology was operating.
The launch follows the force’s announcement in July that facial recognition camera vans would be deployed across Nottinghamshire.
The Wire captured this video and pictures of the technology operating in Old Market Square during its first deployment.
Aug 22, 2026 · via westbridgfordwire.com
From Fuzozo to Pophie: What AI Companion Devices Teach Us About Building the Next Generation of IoT Products Lawrence Wu Lawrence Wu For decades, connected devices have been built around commands. A user presses a button. A sensor reports a reading. A mobile app sends an instruction. Even voice assistants largely followed the same pattern by waiting for a wake word before processing a request. A new generation of AI-native devices is changing that interaction model. Products like AI companions, consumer robots, assistive technologies, educational devices, and wellness products are expected to hold natural conversations, remember previous interactions, recognize different users, respond emotionally, and stay available throughout the day. Instead of reacting to isolated commands, they participate in continuous interactions. That seemingly small shift fundamentally changes the engineering requirements behind the device. Building these experiences is no longer just about integrating a language model. Developers must solve a new class of real-time systems problems involving latency, speech recognition, interruption handling, identity, memory, synchronization, and global infrastructure. Two recent products illustrate this particularly well: Robopoet's Fuzozo and InsBotics' Pophie. Although they target different audiences, they reveal a broader lesson about where IoT architecture is heading. When users are speaking naturally, every pause becomes noticeable. Conversations involve overlapping speech, interruptions, changing speakers, emotional tone, background noise, and long-running context. Unlike issuing a command to a device, conversation is continuous and highly sensitive to timing. The engineering challenge shifts from transmitting data efficiently to maintaining a believable interaction. Instead of optimizing only network throughput or cloud connectivity, developers must optimize something much harder: the flow of human conversation. Most engineers understand that lower latency improves responsiveness. For conversational devices, latency affects something deeper. It shapes whether an interaction feels natural. People instinctively expect conversations to flow without awkward pauses. Delays that might be
Aug 22, 2026 · via iotforall.com
Finds documents with both search terms in any word order, permitting "n" words as a maximum distance between them. Best choose between 15 and 30 (e.g. NEAR(recruit, professionals, 20)). Finds documents with the search term in word versions or composites. The asterisk * marks whether you wish them BEFORE, BEHIND, or BEFORE and BEHIND the search term (e.g. lightweight*, *lightweight, *lightweight*). Lightweight neural networks are increasingly used for image classification in scenarios where memory, computation, and energy resources are limited. However, reducing model size often results in a noticeable loss of robustness and generalization, as compact architectures struggle to learn fine-grained structural cues such as edges, textures, and high-frequency patterns. This limitation is especially evident in domains like medical imaging, microscopy, and remote sensing, where subtle spatial variations play an important role in discrimination. In this work, we introduce LiFGANet, a lightweight frequency and gradient aware network designed to mitigate this problem by incorporating explicit inductive biases into feature learning. Instead of increasing model capacity, LiFGANet emphasizes structurally informative representations through an efficient frequency gradient interaction mechanism, combined with sparse feature refinement and multi-stage feature aggregation. The proposed design remains compact, with approximately 1.13M parameters, and introduces negligible inference overhead. We evaluate LiFGANet on PneumoniaMNIST, BloodMNIST, FashionMNIST, and EuroSAT. The results show consistent and well-balanced performance across datasets surpassing state-of-the-art lightweight models with similar or larger parameter budgets. Ablation studies further demonstrate the various advantages of frequency guidance and gradient modulation. The code and implementation details are available on GitHub. Show AI generated summary Abstract Dive into the groundbreaking LiFGANet architecture, a lightweight neural network designed to revolutionize image classification in resource-constrained environments. This article explores how LiFGANet tackles the fundamental limitations of compact models by embedding physics-inspired inductive biases that explicitly guide learning toward structurally meaningful features. Discover how
Aug 22, 2026 · via springerprofessional.de
Finds documents with both search terms in any word order, permitting "n" words as a maximum distance between them. Best choose between 15 and 30 (e.g. NEAR(recruit, professionals, 20)). Finds documents with the search term in word versions or composites. The asterisk * marks whether you wish them BEFORE, BEHIND, or BEFORE and BEHIND the search term (e.g. lightweight*, *lightweight, *lightweight*). This study presents a novel approach to classify five skateboarding tricks (Kickflip, Frontside-180, Nollie Frontside Shove-it, Pop Shove-it, and Ollie) using transfer learning models integrated with Support Vector Machine (SVM) classification. As skateboarding continues to gain prominence in competitive sports, including its Olympic debut, there is increasing demand for objective evaluation systems. The methodology captures skateboarding trick sequences using a YI action camera positioned 1.26m from the performance area and extracts image frames at 30fps. By overall of approximately 750 images were extracted and then would proceed through a train, validation, and test split of 60:20:20 ratio, respectively. Four pre-trained CNN architectures (NasNetLarge, NasNetMobile, MobileNetV2, and MobileNet) were evaluated as feature extractors coupled with SVM classification. Comprehensive evaluation revealed that NasNetLarge achieved the highest classification accuracy of 93% on the test dataset, followed by NasNetMobile (92%), MobileNetV2 (91%), and MobileNet (87%). Confusion matrices indicate specific patterns of misclassification between similar tricks. This objective evaluation system provides a foundation for automated trick recognition in competitive skateboarding, offering potential applications for objective judging systems in competitions and as a training tool for skateboarders seeking performance improvement. Show AI generated summary Abstract This chapter presents a groundbreaking method for classifying skateboarding tricks using advanced image processing and machine learning techniques. The study begins by outlining the growing popularity of skateboarding and the need for objective evaluation methods, especially as the sport gains Olympic recognition. Researchers developed a three-phase experimental setup: capturing trick
Aug 22, 2026 · via springerprofessional.de
Debate sparks among residents in Town of Irmo about flock cameras IRMO, S.C. (WACH) — Some new traffic cameras in the Town of Irmo are causing a bit of controversy for some people who live there. The cameras are making a noticeable presence in the town, as well as across the state. "If it's being used for other means, you know, following people or things outside of that, I really don't see a use for it," David Cope, Lexington County resident, said. Town residents have been showing mixed emotions after Irmo officials announced plans to install over 20 new flock cameras. RELATED | Irmo officials address online misinformation after safety camera expansion "This place is pretty mellow, but lets keep it that way," Robert Aguilera, Town of Irmo resident, said. According to the Town of Irmo website, the cameras are strictly to be used as investigative tools to help law enforcement. The cameras take a picture of the rear of a car to read the license plate, and do not use facial recognition technology. Town councilmember Gabriel Penfield is the only town council chairholder who opposes the additional flock cameras. "We have a level population, we have a fully staffed police department, so I was just very curious of the need for identification came from," Penfield said. The Midlands have seen flock cameras cause issues over the last month, Including two cameras in Irmo being vandalized last week and a former Richland County deputy being arrested after misusing the camera system. Penfield says he agrees with the people who call these cameras an invasion of privacy. "It is a lack of consent for having that information collected and then ultimately stored in public company servers," Penfield said. Stephen Miano, owner of that computer store, says after the cameras collect a
Aug 22, 2026 · via abcnews4.com
Abstract Neuromorphic vision functionalities have been realized by coupling quasi-bound states in the continuum (quasi-BICs) to multiple quantum wells (MQWs). The engineered leaky modes enhance infrared absorption and generate coexisting nonlinear and linear photoresponses that support image preprocessing and in-sensor computing. This approach highlights a new role for quasi-BIC leakage in integrated optoelectronic intelligence. Biological vision systems perform substantial information processing before signals reach the brain1. Within the retina, operations such as contrast enhancement, noise suppression and feature extraction are executed locally, reducing the computational burden of higher visual centers and enabling highly efficient perception. Reproducing such front-end processing capabilities in artificial hardware has become a central objective of neuromorphic vision technologies, which seek to overcome the latency and energy costs associated with conventional von Neumann architectures where sensing and computation are physically separated2,3. Among various optoelectronic platforms, multiple quantum wells (MQWs) provide a promising route for infrared neuromorphic vision systems due to their strong quantum confinement effects, fast carrier dynamics, and compatibility with mature semiconductor fabrication technologies4,5,6,7. These structures have been widely explored in infrared imaging, optical communication, and sensing applications, making them attractive candidates for in-sensor visual processing8,9,10. However, the optical response of MQWs is fundamentally governed by intersubband selection rules, which require a dominant out-of-plane electric-field component to efficiently drive carrier transitions11,12. As a result, the absorption efficiency under normal incidence is intrinsically limited, posing a critical challenge for direct integration into compact imaging systems. To overcome this limitation, a variety of photonic coupling strategies have been explored. Metallic gratings were introduced to provide the required out-of-plane electric-field component under normal incidence13, while plasmonic resonators14 and metamaterial absorbers15 were subsequently employed to enhance local electromagnetic fields and improve detector responsivity. More recently, metasurfaces have emerged as a versatile platform for tailoring light–matter interactions in MQWs through engineered
Aug 21, 2026 · via nature.com
India appears to be normalising the technical ability to subject political gatherings to searchable biometric surveillance without first having settled, through legislation and judicial oversight, the circumstances in which the state may lawfully do so. While the Delhi Police have continued to deny the use of excessive force and manhandling people involved in the Cockroach Janta Party protests, contrary to protestors’ testimonies, facial recognition equipment was present at the protest site while the demonstrations were on. According to its affidavit to the Supreme Court of India, the police deployed a facial recognition system (including Artificial Intelligence-enabled cameras to scan faces in real-time against a database), a mobile surveillance van and a command and control vehicle, smart spectacles for identifying individuals on the move, and drones and videographers; the van and the spectacles have been tied to private contractors. The police have also not addressed whether actual biometric processing occurred with every individual within the range of cameras. Even if the police discarded the images associated with infructuous checks, that the data may have been accessible to private contractors, whose terms of engagement are unclear, is worrisome. However, no statute governs the use of facial recognition systems; the Digital Personal Data Protection Act, whose data-processing obligations are not yet in force, still makes broad exemptions for state agencies. The Automated Facial Recognition System of the National Crime Records Bureau can be used to identify criminals, unidentified bodies, and so on, and the Criminal Procedure (Identification) Act 2022 expanded the set of records that the police may collect from specified persons. However, these actions are still only allowed vis-à-vis specific groups of people. Irrespective of the constitutional validity of the police’s actions, the chilling effect on potential participants may also curtail the right to protest. The state has to pass a well-established
Aug 21, 2026 · via thehindu.com
Police have made several arrests and recovered a weapon following the latest deployment of live facial recognition cameras in Bradford city centre. West Yorkshire Police used the technology over three days from August 19 to 21, with support from the Bradford City Neighbourhood Policing Team. The cameras helped locate a missing person and identified four individuals wanted for criminal offences, all of whom were arrested. One of those detained, a man wanted for theft, was also arrested on suspicion of possessing an offensive weapon after officers found a knuckle duster. Inspector Justin Adams of the Bradford City Neighbourhood Policing Team said: "The Live Facial Recognition Team have again supported City NPT not only in helping us locate wanted individuals but also in supporting our commitment to protecting vulnerable people. "It has helped officers find a high-risk missing person quickly, ensuring they received the support and safeguarding they needed at the earliest possible opportunity. "The technology also helps us identify those on orders who may cause harm, enabling officers to intervene, engage and take action to keep our communities safe. "Local businesses and traders have broadly welcomed the use of LFR over the last six months, which is now supported by funding from the Knife Crime Concentration Fund." Since February 2026, live facial recognition camera vans have been deployed in Bradford 17 times, resulting in 29 arrests. The system also flagged 21 individuals subject to Sexual Harm Prevention Orders, who were then checked for compliance. Live facial recognition works by comparing real-time camera images to a watchlist of wanted individuals and suspects. Alerts generated by the system are reviewed by trained officers before any action is taken. Images that do not match anyone on the watchlist are deleted within seconds. West Yorkshire Police said the technology is helping to keep communities
Aug 21, 2026 · via thetelegraphandargus.co.uk
Idemia PS face biometrics score high accuracy at scale in latest NIST FRTE Idemia Public Security’s latest facial recognition algorithm has ranked first across the tested gallery sizes in the frontal mugshot-to-mugshot scenario of NIST’s latest FRTE 1:N Identification evaluation. The submission recorded a 1.11 percent FNIR against the 12 million-person gallery, with an FPIR of 0.1 percent. Its ability to maintain low error rates as the enrolled population increased shows the strongest result. Scale provides useful context for differentiating leading biometric algorithms as they converge on controlled image comparisons. Idemia’s (013) face biometerics algorithm delivered an FNIR of 0.05 percent in mugshot identification with databases of 1.6 million and 12 million images, and 0.55 percent on webcam-to-mugshot comparisons, all best in the field. The company’s results remain close to those of the leaders in other controlled or near-frontal scenarios. In visa-to-border comparisons, Idemia scored a 0.13 percent FNIR, third overall but also tops so far in the “by developer” rankings, because NEC’s (010) submission scored better overall than its (011) and (012) algorithms, which have lowest two FNIRs to date. Idemia ranks ninth for visa-to-kiosk identification at 4.74 percent and twenty-sixth when matching profile images against a mugshot gallery with an FNIR of 7.50 percent. The company also finished in the top tier for demographic fairness by FPIR, according to an announcement celebrating its results. The new submission shows a substantial improvement over its algorithm evaluated last year. In visa-to-border identification, FNIR declined from 0.22 percent to 0.13 percent at the same 0.3 percent FPIR operating point. Progress also extends to several other tests, with border-to-border FNIR falling from 2.42 percent to 1.21 percent and mugshot-to-webcam FNIR declining from 0.82 percent to 0.55 percent. The visa-to-kiosk performance improved from 5.16 percent to 4.74 percent. “These results validate our investments
Aug 21, 2026 · via biometricupdate.com
AI Inference — GPT2 Artificial Intelligence and Machine Learning have enabled us to create applications that are almost magical in their abilities. GPT-2 is a powerful language model developed by OpenAI, the makers of ChatGPT, that excels at generating human-quality text. It's a key AI workload because it tests the efficiency and capability of AI systems in handling complex language tasks, which is crucial for many real-world applications. We are measuring the time it takes to generate 100 stories starting with the "Once upon a time, there was a" prompt. AI Inference — Stable Diffusion Stable Diffusion stands out as the second rock star AI workload. This cutting-edge image generation model creates high-quality, detailed visuals from textual descriptions. Its ability to produce realistic images has wide-ranging implications for industries such as art, design, advertising, and media, enabling innovative content creation and enhancing visual storytelling. We are timing how long it takes to generate a single image for the prompt "a photo of an astronaut riding a horse on Mars." AI / Inference — Image Upscaling Topaz Photo AI is a premier tool for AI-driven image processing. It enhances image resolution and quality by intelligently increasing detail and reducing noise. This advanced capability improves the clarity and sharpness of assets, making it a valuable tool for anyone needing high-resolution images. Our test reports how long it takes to upscale a 1.5 megapixel image to 22 megapixels. AI Training — Natural Language Processing Natural Language Processing (NLP) involves training AI models in understanding and generating human language. By applying algorithms and machine learning techniques to analyze and interpret text data, NLP enables tasks like translation, sentiment analysis, and text generation. We are measuring how long it takes to train a BERT language model with a collection of movie critic reviews. AI Training
Aug 21, 2026 · via techpowerup.com
This week, EFF, along with Big Brother Watch, Defend Digital Me, Liberty, Open Rights Group, Race Equality First, Statewatch, and Stopwatch, wrote to Nottinghamshire Police Force in the UK raising concern about the proposed roll-out of live facial recognition technology (LFR), and called for its immediate halt. In particular, the letter highlights six concerns: LFR Is Not "Just Another Tool" Nottinghamshire Police has stated that “facial recognition is just another tool to fight crime.” But LFR used in public spaces is an incredibly intrusive biometric mass surveillance technology that scans the faces of everyone who walks past the camera and takes biometric face prints. This is not just another tool, but a major escalation of surveillance that treats everyone as a suspect by default. People Having "Nothing to Worry About" Does Not Hold to Scrutiny According to Nottinghamshire Police, “if you aren’t entering the city or county to commit crime then you have nothing to worry about.” However, many people have legitimate concerns about the normalisation of invasive technologies. So a public that cannot move around their towns and cities without being subjected to a biometric identity check may be less willing to seek medical care or legal advice, speak with journalists, act in a union, vote, protest, or express their gender, sexual or religious identity. Disproportionate Targeting With LFR We are particularly concerned to learn that Nottinghamshire Police could deploy LFR to tackle low level crimes, such as youth behavior deemed anti-social, as part of Operation View. Reporting suggests that the force already possesses “a watchlist of young people believed to be causing the most problems,” including children as young as 11 years old. It would be highly disproportionate to deploy live facial recognition to tackle this behaviour. Many of these children are reportedly known to the police, and
Aug 21, 2026 · via eff.org