CardSight AI Supports Identifying Multiple Cards in One Image Custom-trained computer vision meets collectors where they are â identifying card a full binder page, while legacy tools still scan one card at a time We train our identification for the real-world conditions collectors actually shoot in, not lab conditions, and our AI is built to read several cards in a single image. — Signe Bone, Founding Engineer at CardSight AI PORTLAND, ME, UNITED STATES, June 29, 2026 / EINPresswire.com/ -- CardSight AI, the computer-vision platform powering trading card identification for developers, marketplaces, and hobbyists, today launched "Break Out of the Box," a campaign built around a core advantage of its technology: users never have to line a single card up inside a guide box and scan it one at a time, like depositing a check in a banking app. With CardSight AI, a single photo is identified in one shot, whether it's taken at the card show, the local shop, or the table where you sort your collection, at any angle and in any light, even a full nine-card binder page. As competition intensifies, the campaign draws a deliberate line between CardSight AI's purpose-built approach and the generic techniques many competing tools still rely on. The problem with scanning "in the box" Most card-identification tools depend on legacy image-matching techniques such as perceptual hashing (p-hashing) and k-nearest-neighbor (kNN) matching. These methods compare a new photo against a library of reference scans by measuring how visually similar the two images are. That only works when the new photo closely mirrors the reference image, which forces the card to be flat, centered, evenly lit, and captured one at a time. The result is what CardSight AI calls the "deposit-a-check" experience: hold the card inside an on-screen box, keep it still, wait for
Jun 29, 2026 · via tech.einnews.com
New spying threats force rethink of biometric identity checks The Five Eyes intelligence alliance’s warning this month that Chinese intelligence services are using fake recruiters on LinkedIn and other job platforms to cultivate people with access to sensitive information is a reminder that the most consequential security failures often begin before a system ever performs a check. Made up of Australian, Canadian, New Zealand, UK, and U.S. intelligence agencies, their joint bulletin, Safeguarding Our Secrets, describes how Chinese intelligence officers or their affiliates pose as recruiters, consultants, and representatives of credible-appearing companies to identify people with access to government, military, economic, or policy information. Publicly available information can also be coupled to the de-anonymization of information obtained through data brokers. De-anonymization occurs when data that has been stripped of direct identifiers, such as names or email addresses, is combined with other datasets to re-identify individuals. All of which is designed to turn a real person into an insider. A job offer provides the opening. A remote interview creates rapport, and the request for something like a trial analytical report tests the target’s willingness to provide information. The demands then become more sensitive. But the episode also illustrates a wider problem emerging across digital identity systems. Security may be built around sophisticated facial recognition, document authentication, and biometric matching, yet these tools can do only so much when the person, the credential, or the digital stream reaching the system has already been manipulated. A biometric scanner can compare a face with a stored image, determine whether a fingerprint resembles a template on file, and can confirm that the person in front of a camera resembles the person associated with a passport or account, but what it cannot automatically establish is whether the identity entered into the system was genuine at the
Jun 29, 2026 · via biometricupdate.com
Artificial intelligence (AI) gained significant public attention when IBM's Deep Blue defeated the world chess champion in 1996 and 1997. In 2011, IBM's Watson surpassed human champions on Jeopardy, illustrating the integration of curated human expertise into computer systems through the analysis of thousands of grandmaster games and consultation with chess professionals and Jeopardy winners. By 2022, generative AI software based on foundational models had become prevalent, supported by increased computing power and extensive training data derived from photos, comments, and captions shared on social media. This growth in software capabilities enabled diverse AI applications, including: - Natural language processing - Image recognition - Face recognition - Autonomous driving - Speech recognition - Robotic AI Millions of mobile phone users depend on AI-powered software for various tasks. Virtual and voice assistants such as Alexa, Siri, and Gemini are widely utilized for daily activities both inside and outside the home, including remote operations. During travel, individuals encounter AI technologies used by public officials for image and face recognition. Employees in manufacturing facilities, regardless of size, also employ classical AI within controlled production environments. A recent Microsoft report indicates that artificial intelligence usage is increasing annually. The latest Microsoft Education survey, which included grade schools through universities, found that 90% of education leaders, educators, and students have used AI at least once for school-related purposes. The report further notes that institutions are leveraging AI to enhance learning and provide recommendations to assist educators and administrators. Notable examples include the Catholic University of Chile in Santiago, which deployed 194 AI pedagogical agents to support learning and improve the teaching experience. Among the most active agents, students averaged 13.2 minutes of sustained engagement, suggesting in-depth exploration of course content rather than superficial queries. In Broward County, Florida, the district implemented 20,000 Microsoft 365 Copilot
Jun 29, 2026 · via blackengineer.com
It’s not uncommon to feel pangs of regret after a big weekend out — for overindulgences of drinks, dalliances, burritos, or all of the above. Now some hungover bargoers can add “turned over personal information to surveillance companies” to the list. A number of destinations in the Castro, including The Mix Bar, Badlands, and Toad Hall, have been using the third-party security service Patronscan to scan IDs, photograph faces, and store personal information in a shared database, as first reported by Gazetteer SF. (opens in new tab) The extra level of security is meant to give businesses a way to flag problematic patrons and share that information with other bars — a kind of mass joint “86’d” list. However, the technology has concerned community members and privacy organizations. “We advise San Franciscans avoid such bars until they remove the facial recognition technology to ensure safety for the queer and trans community, free from harmful surveillance,” digital rights advocacy group Fight for the Future wrote in a post (opens in new tab) ahead of Pride weekend. The good news is that California’s privacy laws, stronger than those of most states, give you some control over what happens to that information. You can see it, fix it, delete it, and, if you feel you were incorrectly flagged by a bar using the software, dispute it. What they have and where it’s shared Before you go into a bar, look around the security check. Is there a small camera pointed at you or a posted disclosure about a scanning system? If they are using Patronscan or similar technology, you can ask about opting out, but the bar is within its rights to refuse you entry. “It’s a really unfair choice, especially for people in more vulnerable populations,” said Hayley Tsukayama, director of state
Jun 29, 2026 · via sfstandard.com
BuzzFeed GamesIf You Can Solve This Color Puzzle In Less Than 3 Minutes, You Have Perfect Color VisionHuedoku #54! New week, new puzzle — let’s see what you’ve got. 🌈🧩Posted 10 hours agocommentFacebookPinterestLinkby Crystal RoBuzzFeedBuzzFeed StaffHi, I’m Crystal, a Senior Editor based in Los Angeles and creator of BuzzFeed’s “That Got Dark” newsletter. Huedoku is a daily color puzzle with simple rules and a satisfying solve. It's like sudoku, but with colors instead of numbers. Sign up here to get notified every time we publish a new Huedoku! Come back tomorrow for Huedoku #55 — and share your score to challenge a friend! 🌈 🌈 New Huedoku drops every weekday at 4:00 a.m. PT / 7:00 a.m. ET. 🧩🗂️ Huedoku Archive — Every past puzzle, all in one place! Comments
Comments
Jun 29, 2026 · via buzzfeed.com
Man arrested by police using facial recognition
- Published
A man was arrested after live facial recognition technology was used in Peterborough city centre.
Cambridgeshire Police used the system in the city for the second time earlier this month.
Of the 22,000 faces scanned on 19 June, two came up as matches against a watchlist, the force said.
The man who was arrested had failed to appear in court on suspicion of driving while disqualified.
The system was first used in Peterborough on 19 May, when police scanned 34,000 faces in six hours.
Two men who were wanted for failing to appear in court were arrested on that occasion â one accused of theft and the other of shoplifting.
Cambridgeshire Police said images of people who did not match the database were permanently deleted straight away.
Do you have a story suggestion for Peterborough? Contact us below.
Get in touch
Your Voice
Follow Peterborough news on BBC Sounds, Facebook, external, Instagram, external and X, external.
Related topics
- Published11 June
- Published19 May
Jun 29, 2026 · via bbc.co.uk
Agencies
Agencies
Audience intelligence
Know your target audienceBuild a complete picture of your customers' complex lives, with audience intelligence from 30 million registered panel members worldwide. Available in dashboards, API, MCP & more.
Know your target audienceBuild a complete picture of your customers' complex lives, with audience intelligence from 30 million registered panel members worldwide. Available in dashboards, API, MCP & more.
Audience intelligence
Know your target audienceBuild a complete picture of your customers' complex lives, with audience intelligence from 30 million registered panel members worldwide. Available in dashboards, API, MCP & more.
Know your target audienceBuild a complete picture of your customers' complex lives, with audience intelligence from 30 million registered panel members worldwide. Available in dashboards, API, MCP & more.
Why YouGovYouGov offers the most accurate data, engaged panel members, pioneering technologies, and trusted experts. You’ll get honest, hard-hitting, real-world data to shape your strategy. Empower yourself with real opinions from real people, tracked over time.
Why YouGov
Why YouGov
Jun 29, 2026 · via yougov.com
Vadzo Imaging Launches AR0521 USB Camera for Interactive Digital Signage and Real-Time Gesture Recognition Applications Vadzo Imaging's Falcon-521CRS is a 5MP USB 3.0 color camera built on the Onsemi AR0521 sensor, delivering low noise rolling shutter imaging with full UVC compliance for interactive kiosk systems, digital signage displays, and real-time gesture recognition applications without custom driver development. FORT WORTH, Texas, June 29, 2026 (Newswire.com) - Vadzo Imaging today announces the Falcon-521CRS, a 5MP USB 3.0 color camera built on the Onsemi AR0521 sensor and designed for OEM engineers developing interactive digital signage platforms, touchless kiosk systems, and vision-based gesture recognition pipelines. The Falcon-521CRS delivers 5-megapixel color imaging at 2592 x 1944 resolution over USB 3.0 with full UVC compliance and a low noise rolling shutter architecture that targets applications where image quality in variable ambient light is a primary engineering constraint. The Engineering Challenge in Interactive Kiosk and Gesture Recognition Systems Interactive digital signage and self-service kiosk platforms are among the most demanding deployment categories for embedded vision camera devices. These systems operate in retail environments, transit hubs, banking terminals, and public information displays where lighting is rarely controlled, and the subject population spans users of different heights, skin tones, and hand positions. A gesture recognition pipeline needs to reliably detect and classify hand shape and motion in real time across this range of conditions without generating false triggers from background movement or light variation. The image sensor sits at the front of this problem. A sensor with poor low-light noise characteristics produces images where hand contours are lost in noise floors below a certain luminance threshold. A sensor with an inadequate dynamic range causes hand tracking algorithms to fail when a user is backlit by a window or a large display panel. A camera module that requires custom driver
Jun 29, 2026 · via newswire.com
The Prospect MAX runs more reliably than most of the fleet, which is why so many people on the east side build their mornings around it. The woman heading to a hospital shift, the teenager riding to school, the man going to dialysis. This fall, if the Kansas City Area Transportation Authority gets its way, every one of them will have their face scanned by an artificial intelligence system the moment they board, and run against a watchlist before they reach their stop. What Kansas City is building, just months or years ago we likely would have described as dystopian. Kansas City wants to be the first, and that is exactly why the rest of the country is watching. “The idea of running face recognition on a camera that is pointed on live spaces in public is a line that until recently has never really been crossed in the last 25 years,” – Jay Stanley, senior policy analyst at the American Civil Liberties Union, told the Associated Press The World Cup Was the Excuse for Scanning Riders’ Faces KCATA tried this once already. It told the public the cameras were about the World Cup, about finding missing people and stopping trafficking during a tournament that draws the whole world to the city. As we know, the tournament is here right now, the matches are being played this month, and the cameras are not on the buses. The reason the agency gave the public (that it is needed for the World Cup) will be over before the cameras go up, yet, KCATA wants to return in the fall with the program more than three times its original size, up to thirty buses. When the program was first announced, the KCATA Board of Commissioners was informed at a finance committee meeting, rather
Jun 29, 2026 · via kansascitydefender.com
Georgia‘s Interior Ministry uses Polyface, a facial recognition system developed by the sanctioned Russian company Papillon AO, to identify and monitor participants in protest rallies, according to the international investigative outlet AlgorithmWatch. According to the publication, the system remains under Russian jurisdiction. It says this increases the risk that Russian security services could gain access to the biometric data of Georgian citizens, posing a threat to both Georgia’s national security and civil activists. AlgorithmWatch notes that Papillon AO is under sanctions imposed by Switzerland, Ukraine, Japan and the United States. According to the publication, Russian law enforcement agencies and countries closely aligned with Moscow, including Tajikistan, Turkmenistan, Kazakhstan and Belarus, primarily use the company’s technology. “Russia, which now supplies Georgia with surveillance technology, invaded the country in August 2008 and subsequently recognised South Ossetia and Abkhazia as independent states,” the publication says. According to the report, authorities have upgraded the Polyface system five times over the past 11 years. In October 2024, they also granted it an indefinite licence. The 2018 upgrade contract required Russian specialists to train operators from Georgia’s Interior Ministry directly. AlgorithmWatch also reports that in early June 2025, Georgia’s Interior Ministry purchased the latest software update, Polyface 3.7.0. The publication says this version relies on an algorithm developed by the Russian company 3DiVi, which is based in Novosibirsk and receives financial support from a Russian state fund. According to the report, the software can capture high-resolution images of crowds numbering in the thousands, even in low-light conditions. It can also identify individuals wearing masks or with partially covered faces. Drawing on procurement documents, AlgorithmWatch also reports that Georgia’s Interior Ministry has removed the existing limit on the number of system operators. Until 2025, no more than 30 operators could use the platform at the same time.
Jun 29, 2026 · via jam-news.net
Pros & Cons - - Slick interface - Useful face recognition tools - Capable auto-corrections - Supports plug-ins and raw files - AI object removal and search - - Available only for Apple devices - Nearly impossible to uninstall on macOS - Weak web interface Apple Photos Specs | Content-Aware Edits | | | Face Recognition | No software is immune to the AI craze, including Apple Photos. It's an indispensable photo editing app that syncs your photos between all your Apple devices and supports Live Photos, Portrait Mode, and the ProRaw format. It comes free with the purchase of any iPad, iPhone, or Mac and offers features you typically see in professional software, including HSL color editing, noise reduction, a vibrance tool, and tone curves. Generative AI remove and powerful search tools are new for the current version, and more AI features are coming later this year. Apple Photos is a fine option for photo hobbyists and a must-download for iPhone users, but the cross-platform Google Photos remains our Editors' Choice winner for entry-level photo editing software. Pricing: Free With the Purchase of an Apple Device The price of an Apple computer, phone, or tablet is all you pay to use Apple Photos, which comes preinstalled. In fact, you can't uninstall it from macOS without taking extreme measures that include command-line operations. Apple's iCloud Photos service, which dependably syncs your photos between iPads, iPhones, and Macs, gives you just 5GB of free storage. Fees range from 99 cents per month for 50GB to $59.99 for 12TB, but the sweet spot for most users is probably the $9.99-per-month 2TB plan. If you want to work on more operating systems, use Adobe Photoshop Elements, Google Photos, or Lightroom instead. For comparison, Google Photos is free for everyone and includes 15GB of
Jun 29, 2026 · via uk.pcmag.com
The Western Australian government is trialling Live Facial Recognition (LFR) technology across the state, with the Australian Retail Council (ARC) welcoming the news. This comes after Kmart and Bunnings were caught up in a review by the Privacy Commissioner over their use of facial recognition technology between 2020 and 2022 to help crack down on retail crime – particularly refund fraud. The ARC called the LFR trialling an encouraging step towards the responsible use of suspect matching technology, with the potential to help protect frontline retail workers and customers from known, high-harm repeat offenders. The WA government did not specify exactly where LFR will be deployed, only noting places like Perth CBD, Maitland and other jurisdictions. “All deployments are overt and will occur in public spaces where LFR has the greatest potential to assist the WA Police Force in fulfilling its operational duties,” the WA government shared. This comes as around 800,000 retail crime incidents were recorded across Australia in 2024, according to the ARC, with one in five of these events involving threats, aggression, intimidation, harassment or other serious behaviours. The peak body added that just 10 per cent of offenders are responsible for around 60 per cent of all retail crime incidents, with repeat offenders up to four times more likely to be violent. Independent national polling commissioned by the ARC also shows Australians strongly support the targeted use of suspect matching technology in situations involving genuine safety risks. Eight in ten (81 per cent) support its use to identify individuals who have previously threatened retail staff with a weapon, with 80 per cent supporting its use to identify people who have physically assaulted retail workers or customers. The ARC added that retailers are keen to see the practical lessons from the WA trial extended to responsible suspect
Jun 29, 2026 · via ragtrader.com.au
Physical AI: Bridging Digital and Real Worlds Physical AI is gaining traction by integrating intelligent agents into real-world environments, enhancing tasks through environmental interaction and motion control. Industry experts foresee significant commercialization potential in various sectors. Par Yu Sinan, People's Daily Artificial intelligence continues to evolve remarkably—from image recognition and text generation to video creation—demonstrating increasingly sophisticated capabilities. As these digital capabilities mature, the technology sector is shifting focus toward integrating AI into physical environments. This emerging concept, known as physical AI, is gaining significant traction within the industry. Physical AI represents intelligent agents capable of perceiving physical environments and performing human-like actions beyond digital interfaces. Ma Xiaojian, head of the joint laboratory between the Beijing Institute for General Artificial Intelligence and Delta Intelligence, noted that physical AI has three defining features: its capabilities are built on real-world physical interaction data, it incorporates an understanding of the physical world, and it can be deployed in real-world physical entities. Where generative AI excels in content creation and data analysis, physical AI specializes in environmental interaction and motion control tasks. "While representing different AI dimensions, these domains demonstrate growing convergence," Ma noted. Generative AI's capabilities—including language interpretation, scenario modeling, and automated coding—enhance physical AI's task execution and environmental navigation. Over the past few years, the tech industry has advanced physical AI from core algorithms to ontology engineering through multiple approaches. Ma said that there are three main technical pathways currently used to implement physical AI. The first is the "pre-training and post-training" approach, in which models undergo large-scale pre-training on internet videos, first-person videos, and cross-robot manipulation data before being further refined through teleoperation data, reinforcement learning, or real-world fine-tuning. The second is the "real-simulation-real" approach, which reconstructs real-world geometry, materials, and dynamics into high-fidelity simulation environments, enabling robots to learn through
Jun 28, 2026 · via alwihdainfo.com
This story by Derechos Digitales originally appeared on Global Voices on June 27, 2026. In Latin America, surveillance is rarely presented as what it is. More often than not, it is set into motion on the premise of other promises: safety, efficiency, and order. It thus colonizes increasingly quotidian spaces until it no longer seems strange to us. Before, when a facial recognition camera seized our attention, we even questioned it. Today, it is part and parcel of public transport, mass events, and football stadiums. The advance of these technologies is not occurring as an exception or temporary measure. They are established silently and often without public debate, transparency, or people truly knowing what data is being recorded, who is storing it, or how it can be used later. At best, it is hidden behind individual consent. At worst, it is assumed to be what the people need. In May 2024, over 1.5 million people gathered on Rio de Janeiro’s Copacabana Beach for a free concert by Madonna. It was a night of celebration, but also one marked by large-scale surveillance, as thousands of agents, drones, and facial recognition cameras were deployed in the name of safety. Derechos Digitales documented the incident and included it in a report for the Inter-American Commission on Human Rights’ Special Rapporteur for Freedom of Expression. The concert surveillance was not limited solely to physical space. According to a media report from Brazil, the Rio de Janeiro Military Police also intensified its social media monitoring as part of its strategy for the event. The so-called “cyber patrol” was conducted without a clear legal framework that established any specific limits, controls, or supervision mechanisms. The crowd was not just observed by cameras at the beach; their online posts, comments, and other interactions were also tracked. This
Jun 28, 2026 · via newsgram.com
6ai — bringing thinking back applied intelligence. Think. Don’t Outsource There was an article the other day in The Telegraph titled “Accenture’s crash shows the consultancy racket is finished.” Although I agree with the article, this journalist’s revelation is a bit late. Anyone paying attention already knows this, and the “crash” has many maturation stages before the racket is gone for good. The consulting industry weakens our businesses, infantilizes our governments, and warps our economies, according to the author of The Big Con, Mariana Mazzucato. Big consulting serves neither citizens nor consumers, and it stunts innovation and obfuscates corporate and political accountability, she goes on to say. The consulting industry has pulled off a confidence trick, which, over the many decades, has hollowed out our thinking. The subtitle says: “Smart chatbots have exposed just how shallow much of the industry has become.” Again, I agree with the premise, but “chatbots” in themselves, as the argument for AI replacing much of consulting work goes, are just more hype to sell more AI stuff. There is always change when new automation technology enters the scene, and the consulting industry is facing the same fate as many other industries have before. But it is fair that this is a very powerful automation technology. The new hype is that companies like OpenAI/ChatGPT and Anthropic’s Claude are taking over from consulting. I don’t think so. These models don’t solve problems, make organizations think and get better. They just speed up information processing and generate outputs that are already programmed in the training data. These models tell us nothing new. Claude AI can provide research in seconds that a junior analyst might take hours or days to produce and package coherently in a slide deck. Then the senior partner comes along and charges $1500 per hour
Jun 28, 2026 · via perry-douglas.medium.com
From teen heartthrobs to swoon-worthy leading men, every decade has had its own class of ridiculously good-looking guys. Below you'll find 20 male celebs people couldn't stop crushing on over the years, and all you have to do is name each one. Good luck!
Comments
Jun 28, 2026 · via buzzfeed.com
Instagram
Log In
Sign Up
Jun 28, 2026 · via instagram.com
Every time you search, browse, or interact with an app, you generate data. That data is worth billions to AI companies. But the platforms that collect it keep almost all the value. A new generation of decentralized AI data marketplaces wants to flip that arrangement — using crypto to pay contributors directly whenever their data trains a machine learning model. The mechanics go deeper than a simple "own your data" slogan. There are verification layers, staking systems, privacy constraints, and token economics — and together they decide whether a contributor gets paid fairly or not at all. This piece explains how those systems work, from the ground up. TL;DR - Decentralized AI data marketplaces connect people who own raw data with AI developers who need labeled, verified training sets, and use crypto tokens to handle payments trustlessly. - Contributors submit data, which is verified on-chain or via decentralized oracle networks before a payment is released, removing the middleman platform from the revenue split. - Privacy-preserving techniques like federated learning and zero-knowledge proofs let data be monetized without the raw underlying information ever leaving the contributor's device. - Token economics, including staking, slashing, and reputation scoring, align incentives so contributors submit accurate data rather than junk. - Projects like Kled AI on Solana represent the current frontier, but the model spans multiple chains and several competing architectures. Why AI Companies Need So Much Data And Who Pays For It Today Large language models and image-recognition systems are data-hungry in a way that's hard to overstate. A single training run for a frontier model can consume hundreds of billions of text tokens, millions of labeled images, or years' worth of recorded human behavior signals. That data has to come from somewhere. Today, most of it comes from a handful of routes. Web
Jun 28, 2026 · via yellow.com
The European Space Agency’s (ESA) Euclid space telescope has captured the largest and most detailed visible-light image ever obtained of the Milky Way's galactic bulge, the central region of our galaxy. The image is a mosaic containing more than 60 million stars, as well as nebulae and star clusters. It will allow scientists to confirm the possible presence of exoplanets using a microlensing technique and measure their masses with greater precision. The Power of Euclid Although Euclid was designed to observe billions of distant galaxies, its visible-light camera is sensitive enough to resolve individual stars at the center of the Milky Way—a region that is both extremely bright and densely populated—without being overwhelmed by the intense light. On March 23, 2025, Euclid turned its gaze toward the galactic bulge, capturing this enormous image in just 26 hours of observations. The result was remarkable: a mosaic composed of nine separate “pointings” (exposures) by its visible-light camera, each covering an area of sky larger than the full moon. While the quality of Euclid's visible-light images is comparable to that of the Hubble Space Telescope, there is one major difference: Each pointing that Euclid captures in just a few hours covers an area 270 times larger than Hubble's field of view. It is also much faster. To put this into perspective, the Keck Observatory would require roughly 2,000 hours to observe the same mosaic. The Image of the Milky Way The new Euclid image captures more than 60 million stars, along with nebulae and star clusters, in one of the Milky Way's most crowded regions—a location ideally suited for searching for exoplanets through gravitational microlensing. “To catch microlensing, you need to observe parts of the sky that are crowded with stars, such as close to the centre of our galaxy,” said Jean-Philippe Beaulieu,
Jun 28, 2026 · via wired.com
Mouse brain activity was used to recreate 10-second videos, offering a new way to study how vision is represented in the brain. Scientists led by University College London (UCL) have reconstructed videos using only brain activity recorded from mice, allowing them to recreate what the animals were seeing. The findings, published in eLife, could help researchers better understand how the brain handles visual information and may offer new ways to study how different species experience the world around them. In recent years, scientists have become increasingly interested in how the human brain makes sense of signals from the eyes. Researchers have shown images and movies to people in fMRI scanners and have tried to decode visual information in the brain down to the pixel level. The new work follows that same broad goal, but it uses single-cell recordings in mice instead. This approach can provide a more detailed view of how the brain represents visual scenes. Using activity from the visual cortex alone, the team was able to produce high-quality reconstructions of videos the mice had watched. Lead author Dr. Joel Bauer (Sainsbury Wellcome Centre at UCL) said: “We wanted to have a better way of investigating how the brain interprets what we see. The current methods of understanding what specific groups of neurons are representing are not very generalizable to situations that haven’t been specifically tested for. And so, we wanted to develop a method that can capture what is being represented in the brain and compare that to reality.” Neurons recreate visual scenes The method could help scientists examine the gaps between what is actually shown and how the brain represents it. Those differences may reveal how particular visual cues influence neural representations. Dr Bauer and colleagues used a dynamic neural encoding model that had been developed by
Jun 28, 2026 · via scitechdaily.com