Member-only story No Finding Is an Island C²A: co-occurrence aware class attention for multi-label chest X-ray classification, and why global average pooling is the wrong last step TL;DR: Standard chest X-ray classifiers end with global average pooling, which flattens the entire image into one shared summary vector and then scores every disease from it. That step destroys two things a radiologist relies on: where a finding sits on the film, and which other findings accompany it. C²A, presented at IEEE MLSP 2026, replaces that step with a classification head that pools separately for each disease under its own learned attention map, then lets the five resulting descriptors exchange information over a graph whose edges are warm-started from the empirical co-occurrence in the training labels. One bounded message-passing step is enough, and the resulting score decomposes algebraically into the evidence a disease found in its own region plus one printable term per related finding. On CheXpert the head reaches 0.895 macro-mean AUROC against 0.8874 for the strongest baseline, adds 0.01 GFLOPs and 0.23 milliseconds per image, and recovers 1.48 points on atelectasis, the one disease where the leading baseline actually regressed below CBAM. Ablations show both halves are load-bearing: removing the graph gives 0.889, and initializing it randomly instead of from…
Aug 29, 2026 · via medium.com
Meta Opens "World's Most Inconvenient Convenience Store" in Shibuya to Showcase Everyday Generative AI Use Through Hands-On Experience US tech giant Meta kicked off a limited-time experiential event on August 28 in Tokyo's Shibuya district that deliberately recreates a difficult shopping environment: "The World's Most!? Inconvenient Convenience Store." The setup challenges visitors to overcome obstacles—staff who don't speak Japanese and unreadable product packaging—using the company's generative AI service "Meta AI" and AI glasses' translation features, with the aim of weaving AI into everyday life. The event is scheduled to run for three days starting August 28, marking Meta's first consumer-facing experiential initiative in Japan. Visitors navigate the store while operating Meta AI installed on their own smartphones. Onigiri rice balls on the shelves feature special labels that make the flavor text illegible; when visitors photograph the packaging and have the AI analyze it, the system displays "kelp flavor." The venue also includes an exhibit where participants generate ad-style images from their own photos and present them to a camera installed on the shelf to unlock a product case. At the checkout counter, staff interact with customers exclusively in Spanish, and visitors can experience real-time translation by wearing AI glasses. Masahiro Ajisawa, Representative Director of Meta's Japan subsidiary, explained: "Meta AI has been available in Japan since November of last year, but we want people to experience firsthand how accessible it is—how Meta AI can support their convenience in daily life." He noted that by inverting the highly convenient everyday space of a convenience store, the company aims to intuitively convey AI's usefulness. The event also features experiences where participants decode product labels composed of Morse code and numerical sequences, allowing them to test AI's core functions—translation and image recognition—throughout a complete shopping journey. Targeting Penetration into Daily Life Meta began
Aug 29, 2026 · via finance.biggo.com
Ring is rolling out TAKE, a new default encryption system. It deletes Amazon's copy of your camera keys within 24 hours - and it's built for one awkward reality: people want AI camera features, but they don't want Amazon holding the keys forever. Amazon's Ring announced the change on August 26. It's replacing its default video encryption with TAKE, short for Throw Away the Key Encryption. According to TechCrunch, the phased rollout starts in September 2026 and TAKE will become the default encryption standard worldwide once it reaches all customers. The pitch is plain enough: Ring wants you to trust it with front-door footage again, without switching off the cloud features that made the camera useful. Here's how it works. Ring says TAKE creates unique encryption keys that rotate every five minutes. A temporary copy sits inside an AWS Nitro Enclave, a cloud environment designed to isolate sensitive processing, so Ring can run features such as Smart Alerts, video search and video descriptions. Then the copy is deleted within 24 hours. After that, Ring says only you and the people you've shared access with retain the keys needed to unlock the video. The company didn't invent the whole cryptographic base from scratch. Amazon's whitepaper says TAKE is built on Messaging Layer Security, an open standard from the Internet Engineering Task Force. That detail matters because a privacy system built only on a company's own black-box promises asks too much of the customer. You still have to trust Ring's implementation, but at least the foundation is a standard security researchers can understand and test. Full end-to-end encryption remains available. It just isn't the default, and for most Ring users that has always been the catch. Ring announced video end-to-end encryption in 2020 and rolled it out more broadly in 2021, but
Aug 29, 2026 · via startupfortune.com
Abstract Angiogenesis plays a crucial role in tumor development as well as in wound healing and integration of biomaterials. However, quantification of complex vascular networks across different in vivo models is time-consuming, labor-intensive, and prone to human error. Deep learning-based analysis (DLBA) appears to be a more efficient and reproducible method for vascular quantification . Yet, studies directly comparing its performance with human assessment remain limited. IKOSA CAM is a commercially available tool built for analysis of vascular networks. Here we evaluated IKOSA CAM against manual vessel tracing across three different angiogenesis models. Images obtained from three different in vivo angiogenesis models, the chorioallantoic membrane (CAM) assay, the rodent retina, and the dorsal skinfold chamber (DSC), were analyzed. The vascular network was analyzed using both manual tracing and the commercial deep learning-based image analysis tool IKOSA CAM. In addition, tracing results were reevaluated by human raters and the total workflow duration was assessed. Deep learning-based image analysis proved to be a time-efficient method for tracing and quantifying vascular networks. The level of agreement with manual human tracing varied among image types. The highest agreement between the two methods was observed for the CAM assay. The greatest differences occurred in the tracing of DSC branching points, as indicated by a downward trend in the Bland-Altman plots. Repeated analyses with the DLBA produced identical results for all tested angiogenesis models, confirming excellent reproducibility. Standardized vascular quantification remains a major challenge in experimental angiogenesis research. In this workflow validation study, the evaluated DLBA provided reproducible and time-efficient vessel quantification, particularly for the CAM assay, while showing limitations in the analysis of retinal and DSC images. Overall, DLBA represents a promising approach for standardized vascular analysis and may facilitate reproducible, high-throughput image quantification. Nevertheless, further validation using larger and more heterogeneous datasets is warranted.
Aug 29, 2026 · via nature.com
Crime & Safety Carpoolers' Road Rage Led To Main Line Shooting; Then, 6 Guns Vanished: Court Docs A bullet in a car door, facial recognition tech and a duffel bag of missing guns helped identify 2 carpoolers as suspects, court docs say. TREDYFFRIN, PA — A bullet lodged in a car door, facial-recognition software, and a duffel bag of missing guns led to arrests this week of two carpooling coworkers accused of a road rage incident that resulted in gunfire last spring in Tredyffrin, according to court documents Patch obtained. Philadelphia residents Sophay Yem and Intanon Chay were charged with attempted murder after a driver accidentally cut them off while getting onto Route 202 and then left the highway with a bullet in his driver's side door, authorities said of the May 11 incident. There were also objects — possibly glass bottles — thrown at the victim's car while Yem tried to run him off the road with brake-checks, flashing headlights and aggressive merging, according to an affidavit of probable cause written by Tredyffrin Police Det. Sara Dougherty. Find out what's happening in Tredyffrin-Easttownfor free with the latest updates from Patch. The victim reported to police the next day that he believed his car had been shot. In the moment, the victim heard what sounded like glass shattering in his ear while he was halfway down the exit ramp. His vehicle shifted. He later discovered a bullet hole in one of his driver's side doors, Dougherty said. Find out what's happening in Tredyffrin-Easttownfor free with the latest updates from Patch. The victim said he "may have inadvertently" cut off a speeding Jeep while entering the Route 30 bypass on the route, according to the affidavit. As both vehicles continued on Route 202 northbound, the Jeep driver pulled up next to
Aug 28, 2026 · via patch.com
in brief - Experts warn that expanding AI surveillance in Australian policing poses serious civil liberties risks. - Privacy advocates have called for stronger laws to prevent unchecked algorithmic overreach in public spaces. Artificial intelligence is increasingly used by police departments across Australia, with experts warning that lines must be drawn over the technology's use, invoking the dystopian world depicted in George Orwell's novel Nineteen Eighty-Four as a potential future to avoid. Police departments are starting to trial the use of AI technology for "various forms of surveillance" and "predictive policing", professor Toby Walsh, chief scientist at the UNSW AI Institute, told SBS News. In WA, police have been trialling AI facial recognition software to scan pedestrians on the street and compare their faces against watchlists. In the trial's first week, 130,000 people were scanned, leading to 18 arrests. Walsh said that while most people would be "pleased" with the result, he believes such public surveillance is "an erosion of the things that we used to value". "The question is, how much are we willing to trade those values for the benefits that the technology is bringing?" News that makes sense Your trusted source for staying up-to-date with the world around you. Get free daily news updates and analysis, straight to your inbox. Overseas, the rise of AI in surveillance technology has become an increasingly contested topic. In the United States, citizens have been taking action against the surveillance cameras of the AI company Flock Safety, which has tens of thousands of devices across the country. Videos of people destroying the cameras have gone viral on social media. Tom Sulston, head of policy at the advocacy group Digital Rights Watch, told SBS News that, while Flock cameras are not currently used outside of the US, "plenty" of companies in Australia
Aug 28, 2026 · via sbs.com.au
Table of Contents That $17B Meta settlement comes with a catch: Your online speech rights Meta agreed on Wednesday to pay up to $17.1 billion to settle lawsuits brought by a bipartisan coalition of attorneys general from 47 states and several territories. The lawsuits, which accused the company of harming young people through the design of their speech platforms, relied on untested legal theories that faced serious First Amendment challenges. Nonetheless, as part of the settlement, Meta agreed to a series of design changes for minors, including a two-hour daily limit, mandatory pauses to interrupt scrolling, blackouts between midnight and 6 a.m., the elimination of visible like counts, and more robust age verification tools. Some are cheering the settlement. But Americans should understand the consequences: These features cannot be implemented without adults having to prove their identity to access protected speech. That could mean document verification, facial recognition, or other invasive hurdles. Courts have repeatedly blocked online age verification laws for social media platforms because forcing people to prove their age before accessing constitutionally protected speech burdens First Amendment rights. On top of that, it’s not just adults who have First Amendment rights — kids and teens have strong First Amendment protections too. This settlement effectively gives government officials a backdoor to achieve through litigation what courts have prevented them from achieving through legislation: the evisceration of anonymous speech rights on a major platform. Anonymous speech has long been a feature of the free and open internet and of American political discourse since the nation’s founding. Don’t be fooled: This is unlikely to stop at Facebook and Instagram. Meta is already urging TikTok and YouTube to adopt the same standards. It is also supporting legislation that would mandate age verification at the app-store level, which would require every user —
Aug 28, 2026 · via fire.org
You must activate javascript to visualize the Virtual Tour : A19610122000 Grumman F4F-4 Wildcat (Virtual tour generated by Panotour).
Aug 28, 2026 · via airandspace.si.edu
Bratislava, Slovakia, August 28th, 2026, Chainwire altFINS j.s.a., a leading provider of proprietary crypto data analytics and AI-driven trading tools, today announced a strategic partnership with BitPath Holdings, Inc. aka Bio-Path Holdings, Inc. (OTC: BPTH) to launch Gen 2 Digital Asset Treasury (“Gen 2 DAT”) strategy. Under the partnership, altFINS will provide institutional-grade crypto analytics, market intelligence, and active portfolio management infrastructure to support BitPath Holdings's corporate treasury pivot. Additionally, altFINS Founder and CEO Richard Fetyko has been appointed to BitPath Holdings' Advisory Board to spearhead its digital asset strategy. Redefining Corporate Treasury Management The partnership marks a significant evolution from traditional "buy and hold" digital asset treasuries (Gen 1 DATs) toward an active, multi-asset management model. While early pioneers demonstrated that cryptocurrencies belong on corporate balance sheets, single-asset strategies frequently expose enterprises to severe concentration risk and market volatility. BitPath's Gen 2 DAT is designed to outperform market cycles through active risk management—utilizing long and short positions and a diversified basket of projects at the intersection of blockchain and AI, including Layer 1 and Layer 2 networks, decentralized finance (DeFi), DePIN, and Real World Asset (RWA) tokenization. To drive performance, altFINS will manage BitPath's crypto exchange accounts and deliver weekly performance reports and strategic guidance. The strategy merges technical price trends with rigorous fundamental token metrics, adapting dynamically to market cycles and sentiment. Strategic Alignment and Transaction Terms To firmly align the long-term growth and success of both organizations: Richard Fetyko, Founder and CEO of altFINS, stated: "The era of passive 'buy and hold' digital asset treasuries is evolving. While early pioneers proved that cryptocurrencies belong on corporate balance sheets, a Gen 1 single-asset approach exposes companies to unnecessary concentration risk and market volatility. Through our partnership with BitPath Holdings, we are setting a new standard for Gen 2
Aug 28, 2026 · via markets.businessinsider.com
Councils use facial recognition in AI cameras to target violence North Tyneside and Preston councils have more than 100 of the CCTV cameras and are scanning for faces in real time to detect aggressive behaviour Previous Article Next Article
Aug 28, 2026 · via thetimes.com
Former Georgia officer arrested over alleged Flock system misuse A former Covington Police Department officer has been arrested after investigators accused him of accessing a police license plate reader system for a non-law enforcement purpose. The Georgia Bureau of Investigation announced Thursday that Michael Canty, 29, of Covington, was arrested Aug. 26 and charged with misuse of a license plate reader system and violation of oath of office. Here's what to know. What investigators say happened The Covington Police Department asked the GBI to investigate on Aug. 6 after allegations surfaced involving the department's Flock Safety license plate reader system. According to the GBI, preliminary information indicates Canty accessed the system for a purpose unrelated to law enforcement. The GBI did not provide additional details about what information was accessed or how the system allegedly was misused. Canty is no longer with the department Canty is a former Covington Police Department officer. Following his Aug. 26 arrest, he was booked into the Newton County Jail. The GBI said the investigation remains active and ongoing. Once investigators finish their work, the case will be turned over to the Alcovy Judicial Circuit District Attorney's Office for prosecution. How to share information with investigators Anyone with information about the investigation can contact the GBI Tipline at 800-597-TIPS (8477), submit information online through the GBI, or use the See Something, Send Something mobile application. How Flock cameras work Flock Safety, an Atlanta-based public safety technology company, develops the cameras used in communities across the country. When a vehicle passes a camera, the system captures an image and converts details about the vehicle into searchable data. Law enforcement agencies can then use that information during investigations, such as locating stolen vehicles or identifying vehicles connected to reported crimes. Flock says the cameras are designed to
Aug 28, 2026 · via augustachronicle.com
Verifying your device, please wait
Verification failed. Please try again.
Try Again
Verification failed. Please try again.
Aug 28, 2026 · via thesun.co.uk
Standing at the TSA podium, a camera points at your face. The CAT-2 kiosk captures your live image and matches it against your passport or ID photo in real time—automating what used to be a purely manual check. TSA frames this as faster and smarter. Privacy experts call it something else entirely, raising concerns about surveillance technology that may extend well beyond its stated purpose. Here’s what TSA says about the program: - Deployed at 80+ airports; planned expansion to 400+ federalized airports - Officially optional—travelers may decline “without recourse” - Photos not stored after a successful match, except in limited testing environments - Manual ID check remains available; opting out won’t cost you your place in line - Framed as identity verification only—not surveillance or law enforcement Why Experts Say Opt Out Anyway Civil liberties advocates and bipartisan lawmakers are raising alarms that go well beyond convenience. The Electronic Privacy Information Center isn’t subtle about its position: “Facial recognition is an invasive and dangerous surveillance technology… TSA’s use of facial recognition normalizes the use of our face as our ID, and the dangerous implications are far too great to ignore.” Optional today doesn’t guarantee optional tomorrow—and cases of apps secretly tracking users show how quickly stated limitations can erode. Senators Jeff Merkley and John Kennedy—a Democrat and a Republican—introduced bipartisan legislation to restrict the program. When those two agree on something, that’s worth your attention. Then there’s the accuracy problem. NIST’s Face Recognition Vendor Test found false positive rates 10–100 times higher for Asian and African American faces compared to Caucasian faces. Lead researcher Patrick Grother documented “demographic differentials in the majority of the face recognition algorithms” studied. For Black women specifically, one-to-many false positive rates were among the highest recorded. It’s a bit like a Spotify algorithm confidently
Aug 28, 2026 · via gadgetreview.com
Live Facial Recognition (LFR) vans are set to come to the Eden Centre next week.
Police have said that the LFR vans will be at the shopping centre next Friday, September 4.
The force said specialist LFR team will working alongside local officers to identify known suspects, deter crime and help keep communities safe.
A spokesperson for Thames Valley Police said: "We make sure to let you know before our LFR vans are in use, so everyone stays informed about what’s happening.
Read more:
High Wycombe man fined following incident in Marlow
Special constabulary officers carry out fail to appear warrants operation
Appeal launched after person attempted to gain entry to Beaconsfield property
"While this can mean some individuals choose to avoid us, the vans can help deter crime and keep the area safer.
"Preventing crime is just as important as catching those who break the law.
"If you have any questions or would like to see how the technology works, feel free to come over and chat with our team.
"We’re always happy to show you around and explain how LFR is used fairly and responsibly."
Share
Aug 28, 2026 · via bucksfreepress.co.uk
Key Points - Apple is rolling out its iPhone 18 family of products. - Analysts are also expecting the company to release its first folding phone. Apple (NASDAQ:AAPL) has been a transformative tech company, completely disrupting the smartphone market by introducing touchscreen controls, internet service, and easy access to texting, music, and games. But dramatic innovations have been far less common in its more recent iPhone models, leading some of the shine to come off Apple. Is that about to change? Apple has scheduled an iPhone launch event for Sept. 9, marking the first such event under the leadership of John Ternus, who is taking over from Tim Cook as CEO on Sept. 1. Missed Nvidia in 2009? This Rare Signal Is Flashing Again.In 2009, a "Double Down" signal flashed for a little-known chipmaker called Nvidia. For the first time in years, that same "Total Conviction" signal is flashing for a company 1/100th the size of Nvidia. Continue » Ternus' first launch event could be huge -- and include a product that would be Apple's biggest phone change since it rolled out its iPhone X and 3D facial recognition. Apple incoming CEO John Ternus. Image source: Apple. The folding iPhone may finally be here The Sept. 9 event is widely expected to focus on the iPhone 18 family of products, including the iPhone 18 Pro and iPhone 18 Pro Max. That will likely also include the new 2-nanometer A20 Pro chip, which is expected to run cooler and longer, and a C2 chip that would provide better battery efficiency, enhanced privacy, and improved performance in congested data networks. But if you're looking for a bigger physical change, you'll likely be more interested in the company's upcoming iPhone Fold (or possibly called iPhone Ultra), which will be Apple's first folding smartphone.
Aug 28, 2026 · via theglobeandmail.com
Online casinos have long relied on passwords and PIN codes to protect player accounts, but these traditional methods are increasingly vulnerable to hacking, phishing, and simple human forgetfulness. As security threats grow more sophisticated, operators are turning to biometric authentication to offer a safer and smoother experience. Fingerprint scanning and facial recognition allow players to access their accounts in seconds using nothing but their own unique physical traits. This shift reflects a broader move across finance and technology toward frictionless, identity-based security. Understanding how these systems work helps players make informed choices about where and how they gamble. Why Casinos Are Adopting Biometric Authentication The gambling industry handles enormous volumes of sensitive data, from personal identification to payment details and transaction histories. Protecting this information is both a legal obligation and a competitive advantage, since players naturally gravitate toward platforms they trust. Biometric login reduces reliance on passwords, which are frequently reused, weak, or stolen in data breaches. It also streamlines the user journey, eliminating the friction of remembering complex credentials. For operators, fewer account recovery requests translate into lower support costs and higher retention. Beyond convenience, regulators in many regions are tightening requirements around responsible gambling and anti-money-laundering compliance. Biometric verification supports these goals by ensuring that the person accessing an account is genuinely its owner. This makes it harder for underage users or fraudsters to slip through the cracks. How Fingerprint and Facial Recognition Work Biometric systems capture and analyze unique physical characteristics, then compare them against a stored template to confirm identity. Fingerprint scanning maps the ridges and valleys of a fingertip, while facial recognition measures dozens of points across the face, such as the distance between the eyes and the shape of the jawline. Modern smartphones already include the hardware needed for both methods, making integration
Aug 28, 2026 · via ritzherald.com
Your face is public. Your permission to use it is not. If someone lifted your photograph and built an account around it, an X face search will surface that account faster than any manual method available to you. You upload a clear photo of yourself, and a facial recognition tool scans social platforms for profiles carrying your face, including accounts you have never seen and would never have thought to look for. Impersonation is rarely discovered on purpose. A friend forwards a screenshot. A stranger sends an angry message about a conversation you never had. The financial scale explains why fraudsters bother. The FTC reported $2.1 billion in losses to scams that began on social media in 2025, an eightfold rise since 2020. How Widespread Is Impersonation on X? Widespread enough that X maintains a dedicated policy and reporting route for it. X’s authenticity policy prohibits two distinct things. The first is impersonating an individual, group or organisation in order to deceive. The second is the use of fake personas, which the policy defines to include stock, stolen or AI generated profile photos, copied biographies, and misleading profile information used to deceive others. A stolen photograph triggers both. X presents particular difficulty for a second reason. Verification was restructured into a paid subscription, which means a blue checkmark confirms a payment rather than an identity check. How Does an X Face Search Actually Work? It compares faces rather than files. That single distinction is why an X face search finds accounts that file-based methods miss. Why Reverse Image Search Misses Most Impersonators A face search and a reverse image search look similar from the outside, but they solve different problems. Upload a photo to Google Images and the engine searches for that specific image: same composition, same crop, ideally the
Aug 28, 2026 · via financialcontent.com
Live facial recognition vans coming to Milton Keynes New police vans that recognise faces will be deployed in Milton Keynes city centre in an effort to deter crime. Thames Valley Police say that live facial recognition (LFR) vans will be coming to Midsummer Place on Saturday, September 5. LFR technology works by analysing key facial features and generating a biometric template. It then compares it to other biometric templates of known faces in the database to find possible matches of people that the police are looking for. A specialist LFR team will be working alongside officers to identify known suspects, deter crime and help keep the community safe. A spokesperson from Thames Valley Police said: “We make sure to let you know before our LFR vans are in use, so everyone stays informed about what’s happening. “While this can mean some individuals choose to avoid us, the vans can help deter crime and keep the area safer. “Preventing crime is just as important as catching those who break the law.” Officers are encouraging anyone who has questions or wants to see how the technology works, to speak to the team on September 5. The spokesperson added: “We’re always happy to show you around and explain how LFR is used fairly and responsibly.” You can find out more about LFR on the Thames Valley Police website.
Aug 28, 2026 · via miltonkeynes.co.uk
How Deepfake Detection Integrates with Modern Identity Verification Systems Digital identity verification is becoming more complex as artificial intelligence makes synthetic images, videos, and identities increasingly convincing. Organizations can no longer depend only on passwords, identity documents, or basic facial matching when protecting sensitive accounts and services. Deepfake detection adds another security layer by analyzing digital media for signs of artificial generation or manipulation. When integrated with facial recognition, liveness detection, document verification, and risk assessment, it can help organizations build stronger identity verification systems designed to address modern impersonation threats. The Changing Identity Verification Landscape Traditional identity verification generally focused on whether a person could provide the correct credentials or a valid identity document. Digital services have changed this process by allowing users to verify themselves remotely. Remote verification creates convenience, but it also introduces new attack opportunities. Fraudsters may use stolen documents, manipulated photographs, synthetic faces, or generated videos to impersonate legitimate users. Deepfake detection helps address the media manipulation component of these attacks. What Deepfake Detection Adds to Identity Verification Deepfake detection examines images, videos, or other digital media for characteristics that may indicate manipulation or synthetic generation. AI models can analyze facial patterns, frame consistency, lighting, texture, movement, and other signals. In video authentication, systems may also examine temporal relationships between frames. The objective is not simply to determine whether content looks realistic. Instead, the system searches for technical patterns that may indicate that the media has been altered or generated. Integrating Deepfake Detection With Facial Recognition Facial recognition and deepfake detection perform different functions. Facial recognition evaluates whether a user's facial characteristics match a trusted identity reference. Deepfake detection evaluates whether the submitted facial media may have been manipulated. Using both technologies creates a more comprehensive verification process. For example, a user may submit an
Aug 28, 2026 · via bignewsnetwork.com
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 28, 2026 · via youtube.com