Image Recognition

KN Group Prevents MX$75 Million Losses With Unico Tech

KN Group Prevents MX$75 Million Losses With Unico Tech Mexico’s fintech sector is accelerating the adoption of advanced identity verification tools—such as the partnership between Unico and KN Group—to counter a 324% surge in account takeover attacks and the rise of AI-driven synthetic identity fraud. The shift is critical for a digital credit market that surpassed MX$450 billion in 2026, enabling institutions to cut operational costs by up to 90% while preserving the rapid origination speeds needed to stay competitive. —— KN Group, a specialist in advanced payment solutions, has identified and blocked approximately 3,000 fraud attempts per month following the implementation of identity validation technology from Unico. By leveraging a system with a detection accuracy rate exceeding 95%, the company projects the prevention of up to 13,000 fraud attempts annually. This represents an estimated economic impact of MX$75 million (US$4.33 million), based on an average credit value of MX$5,000 (US$288). The integration of these security protocols addresses a critical bottleneck in the digital credit ecosystem, where identity theft accounts for roughly 40% of financial fraud in Mexico. Pattern recognition has become increasingly important as serial fraudsters target multiple fintech platforms simultaneously. Unico’s risk engine analyzes 36 behavioral variables to identify “hyperactive” patterns, blocking users attempting to open multiple accounts within short timeframes. “The integration of validation technology has become indispensable for protecting our portfolio,” said Francisco De Hoyos, CEO f KN Group in Mexico. “Each rejected request or high-risk case identified represents a potential loss avoided, whether from fraud-related defaults, recovery costs or reputational damage.” The deployment of automated validation has also improved operational efficiency. According to Fernando Paulin, CEO, Unico, microcredit providers can reduce operating costs by up to 90% by eliminating manual back-office processes and reliance on multiple service providers. For KN Group, this efficiency has supported

Apr 25, 2026 · via mexicobusiness.news

Slashing $15b from NDIS, giving $53b to Defence. See the issue?

In the latest episode of morally vacuous financial gymnastics performed by the Commonwealth, Health Minister Mark Butler spent 90 minutes at the National Press Club on Wednesday justifying more dramatic cuts to the National Disability Insurance Scheme (NDIS). He used every bogus excuse under the sun except the obvious truth: the corporate ruling class simply doesn’t value disabled people, viewing them as economically useless, disposable grifters, gaming the system for free haircuts. Around 160,000 recipients are set to lose access to the NDIS by 2030, with spending forecast to decrease from $70 billion to $55 billion over the same period. Keep reading for free. Create an account to unlock more stories or subscribe to get unlimited access to all articles, Crikey Daily's full edition experience and the early morning Worm newsletter. Support independent journalism and stay informed with a Crikey subscription — no ads, no paywalls, no limits. Keep Reading Create a free account BEST VALUE OFFER Comments

Apr 25, 2026 · via crikey.com.au

Wisconsinites Are Fighting Back Against Flock

Across Wisconsin, something important is happening. Communities are starting to push back against Flock, a rapidly expanding surveillance system that tracks the movement of vehicles through automatic license plate readers. These cameras scan passing cars, log where they were seen, and store that data in a searchable system. Think of it as facial recognition, but for your vehicle and your daily movements. In Wisconsin more than 221 law enforcement agencies, along with some private companies and homeowners associations, use this surveillance system to monitor their communities, often without people knowing. That information does not just stay local. It can be shared across jurisdictions and accessed by outside agencies, including federal law enforcement like ICE and DHS. In some cases, similar data has been used in investigations tied to abortion laws in other states. For years, this system of massive invasion of privacy, expanded with little public awareness or oversight. But that is beginning to change. Communities are finally saying enough is enough and are fighting back. Here are some of the recent wins in Wisconsin against the expansion of this surveillance. Dane County supervisors overwhelmingly voted 31 to 1 to cut funding for Flock cameras in one of the largest counties in the state. This effort was driven in part by local residents like Gehrig Abbattista, who testified at the hearing. He warned that even if the footage is only used in specific cases, its existence creates opportunities for misuse by bad actors. He also pointed out how easily this kind of system could be used to target individuals and vulnerable communities. Another resident, Aedan Gardill, raised concerns about data sharing. While Dane County has said it does not directly share Flock data with ICE, Gardill noted that the county shares data with other jurisdictions that do. Dane County had

Apr 24, 2026 · via aclu-wi.org

Meet the early pioneers who shaped the future of today's self-driving cars

Decades before autonomous vehicles competed in high-speed desert challenges, a small group of engineers in Germany and the United States were quietly building the foundations of self-driving technology. A DRIVERLESS TOMORROW: Explore the triumphs, setbacks and innovations shaping the driverless future Many credit figures, such as Karl Benz, Henry Ford, and Enzo Ferrari, have shaped the modern automobile. But when it comes to self-driving vehicles, a different set of names emerges, including Ernst Dickmanns, Red Whittaker, Sebastian Thrun, Chris Urmson and Anthony Levandowski. Several of these innovators first gained widespread attention during the DARPA Grand Challenges, a series of autonomous-vehicle races launched by the U.S. Department of Defense in 2004 to accelerate progress in self-driving technology. Years before those competitions, however, Dickmanns and his team were already testing autonomous vehicles on public roads. "I'm Ernst Dickmanns, and I was the first who put a self-driving car on the road in the early 1980s," he said. Now nearly 90, Dickmanns lives in a small town near Munich, far from the technology centers of the United States. He rarely speaks publicly about his work, but his contributions helped shape how autonomous vehicles "see" the world. Dickmanns pioneered dynamic computer vision, the technology that allows machines to interpret moving images in real time. In the 1960s and 1970s, he worked with the German Aerospace Research Establishment, analyzing rocket trajectories, applying image-processing techniques at a time when computer power was severely limited. "Well, in Oberpfaffenhofen, I've been the acting director of the Centre in 1974/75," Dickmanns said, describing his early exposure to image evaluation from satellites and aircraft. As microprocessor performance rapidly improved, he concluded that real-time image analysis for vehicles would become possible within his lifetime. By the mid-1980s, Dickmanns and a small engineering team south of Munich converted a Mercedes-Benz van into

Apr 24, 2026 · via abc7news.com

Paciolan Partners With Wicket in Latest Facial Recognition Venue Entry Play

Paciolan is expanding its push into identity-based ticketing with a new partnership with Wicket, introducing facial authentication capabilities to its PAC io Entry platform as part of its broader PAC Link partner ecosystem. The integration will allow colleges, arenas, and performing arts venues using Paciolan to offer opt-in biometric entry, enabling fans to enter venues without scanning a barcode or presenting a mobile ticket. Instead, enrolled users can access events through facial authentication at designated entry points. Paciolan is positioning the Wicket integration as a way to accelerate entry and reduce friction during that critical period. “Paciolan has always focused on providing best-in-class ticketing solutions for our partners and their fans and patrons,” said Brendan Lynch, President of Paciolan.

Apr 24, 2026 · via ticketnews.com

Civilian Quarterly Awards Ceremony 26/1 [Image 26 of 40]

Apr 24, 2026 · via dvidshub.net

Downtown Silver Spring sees sharp drop in crime amid safety push

Crime in downtown Silver Spring dropped significantly in 2025, with sharp declines in gun assaults, carjackings and weapons discharges. Officials credit increased policing, new safety regulations for late-night businesses and expanded use of technology for improving public safety. [The Banner] Misidentified by facial recognition technology, woman wrongfully held for six months Kimberlee Williams was wrongly jailed for six months after facial recognition misidentified her as a fraud suspect in Maryland. The case highlights concerns about the technology’s accuracy and police handling of such evidence. [The Washington Post] Nuclear startup Blue Energy pulls in $380M to scale prefab reactors As power demand surges, Blue Energy is betting that building nuclear reactors in shipyards—rather than on-site—can cut costs and construction timelines. The Chevy Chase startup has raised $380 million to launch its first Texas project using this prefabrication approach. [Washington Business Journal] Today’s weather: Highs around 80, with a chance of rain In case you missed it: Four teenage boys charged in stabbing, robbery near Clarksburg bus stop From Bethesda Magazine: Shibuya Eatery serves Japanese cuisine in a Chipotle style Beyond the box score: Bethesda Magazine co-founder explores top sports stories in new podcast

Apr 24, 2026 · via bethesdamagazine.com

What is a passkey, how does it work and why is it better than a password?

The UK’s National Cyber Security Centre has called time on the password – from now on, you should use a passkey. The NCSC said this week it would no longer recommend using passwords where passkeys were available. They should be consumers’ first choice of login across all digital services because passwords were not secure enough to stand up to modern cyber threats. What is a passkey? Security officials describe a passkey as a “digital stamp” that allows you to sign in to apps and websites and is stored on your device. It is a password-free form of login. Unlike a password, it cannot be stolen in a phishing attack, where people are fooled into handing over their credentials, which can later appear on the dark web. It just requires your smartphone or device to confirm that it is you trying to log in, by using biometric methods such as facial recognition or your phone’s pin. That triggers the “stamp” – or secure passkey – which confirms to the app or website that you are who you say you are. Each account you are registered with will have a different passkey. Even if an app or website using passkeys is breached, it is of no use to an assailant because the device holds the “private” passkey needed to complete a login. Passkeys can also be synced across devices. How do you set up a passkey? The NCSC says you can go to account security or privacy settings on apps and websites you already use, or look out for prompts from services asking you to upgrade to passkeys. You may also be offered to set one up when creating a new account for an app or website. Google says just over 50% of users of its services in the UK have a

Apr 24, 2026 · via theguardian.com

[LINK IN BIO] Read about the new facial recognition technology at Disneyland that takes a photo ...

Instagram Log In Sign Up

Apr 24, 2026 · via instagram.com

3 Stocks to Buy From the Prospering Semiconductor Industry

3 Stocks to Buy From the Prospering Semiconductor Industry Industry Description The Zacks Electronics – Semiconductors industry comprises companies that provide a wide range of semiconductor technologies. Their offerings include packaging and test services, wafer cleaning, factory automation, face detection and image-recognition capabilities to develop smart and connected products. The industry participants primarily cater to end markets that include consumer electronics, communications, computing, industrial and automotive. The companies are increasing their spending on research and development to stay afloat in an era of technological advancements and changing industry standards. The industry is experiencing solid demand for advanced electronic equipment, which is helping its participants increase their investments in cost-effective process technologies. What's Shaping the Future of the Electronics ??? Semiconductors Industry? AI Demand Driving Prospects: Industry participants are benefiting from growing demand for advanced manufacturing processes and energy-efficient computing power, both of which are needed to develop AI-supportive chips. AI is gaining popularity thanks to multimodal learning and growing context awareness. The emergence of Gen AI and Agentic AI has further enhanced AI’s capabilities, making it a key driver of efficiency, automation and innovation. Significant improvements in computing hardware (GPUs and TPUs) are allowing the development of more complex AI models. The growing number of high-speed data centers worldwide, which require ultra-fast Internet that 5G promises to deliver, is a tailwind. Spending on AI infrastructure is expected to accelerate in 2026 as enterprises continue to leverage AI as part of their digital transformation efforts. Smart Devices Aiding Computing Demand: Smart devices need computing and learning capabilities to perform functions like face detection, image recognition and video analytics capabilities. These require high levels of processing power, speed and memory and low power consumption, as well as better graphics processors and solutions, which bode well for the industry. Graphic solutions help increase

Apr 24, 2026 · via theglobeandmail.com

Civilian Quarterly Awards Ceremony 26/1 [Image 22 of 40]

Apr 24, 2026 · via dvidshub.net

From local textures to global attention: a hybrid feature fusion network for skin lesion classification

Abstract Accurate detection of skin cancer detection using RGB images remains a challenge due to multiple factors including variability in lesion appearance, difference in skin types, and the clinical interpretability of the models. To address these challenges, we present a unified feature-fusion framework that integrates deep learning methods for accurate classification of dermoscopic skin lesions. Our proposed method consists of preprocessing in which we performed normalization, class aware selective augmentation, followed by feature-level fusion from three customized deep learning architectures i.e.; VGG-16 with adaptive layer configuration, ResNet-50 with dermatological feature enhancement, and a Vision Transformer (ViT) with dynamic patching. These extracted features are then fused to form a comprehensive feature vector and classification is done using an HDFNet which is a two-layer Deep Neural Network (DNN). We have trained and tested the proposed model on four publicly available datasets including ISIC 2019, ISIC 2020, PAD-UFES, and DermQuest DERMIS. We achieved classification accuracy of 94.5% and an AUC-ROC of 97%. Our proposed method outperforms the existing state-of-the-art models and also provide interpretable predictions supported by Grad-CAM-based visual explanations. Similar content being viewed by others Data availability The datasets generated and/or analyzed during this study are available in the ISIC Challenge repository at https://challenge.isic-archive.com/data/#2019, https://challenge.isic-archive.com/data/#2020, https://data.mendeley.com/datasets/zr7vgbcyr2/1, https://vip.uwaterloo.ca/skin-cancer-detection/. References Mukadam, S. B. & Patil, H. Y. Skin cancer classification framework using enhanced super resolution generative adversarial network and custom convolutional neural network. Appl Sci 13, 1210 (2023). Al-Rasheed, A., Ksibi, A., Ayadi, M., Alzahrani, A. I. & Mamun Elahi, M. An ensemble of transfer learning models for the prediction of skin lesions with conditional generative adversarial networks. Contrast Media Mol Imaging 2023, 1–15 (2023). Thanka, M. R. et al. A hybrid approach for melanoma classification using ensemble machine learning techniques with deep transfer learning. Comput Methods Progr Biomed Update 3, 100103 (2023). Kahia,

Apr 23, 2026 · via doi.org

Disneyland loosens up on controversial tech change to theme park entry

Disneyland offers alternative to controversial tech change to theme park entry Disneyland is offering guests an alternative to a controversial tech feature that is now shaping the way guests enter the park. Now, guests who would rather not hand over a “faceprint” to stroll down Main Street U.S.A. have a choice. Disneyland and Disney California Adventure also include special entry lanes that skip facial recognition, according to the Orange County Register. The entry lets visitors who feel uneasy about sharing their biometric data or fret over cyber risks enter the parks without joining the system, which was originally meant to speed up entry and stop fraud. Both parks offer clearly marked lanes for those who “do not wish to participate” in the service. Disneyland staff may still snap photos of guests in these lanes, but the images skip any facial recognition processing. Kids can also use the opt-out lanes, as long as a parent or guardian gives the green light. Under the standard system, Disneyland photographs daily visitors and annual passholders at the entrance and converts those images into unique numerical values using biometric technology. Those values are then used to match a guest’s face to the photo tied to their ticket or pass. Disney says it deletes those biometric numerical values within 30 days, unless they need to be retained for legal or fraud-prevention reasons, and maintains that it has safeguards in place to protect visitor data. “The security, integrity, and confidentiality of your information are extremely important to us,” the company states, while acknowledging that “no security measures are perfect or impenetrable.” The shift comes as concerns grow about surveillance and tracking in public spaces, giving Disneyland guests a way to skip the tech altogether if they choose. Download The California Post App, follow us on social, and

Apr 23, 2026 · via nypost.com

Jewish man called 'baby killer' in antisemitic attack

Jewish man called 'baby killer' in antisemitic attack Warning: This article contains details that some readers may find distressing A Jewish man has said he feared for his life after being assaulted and called a "baby killer" in an antisemitic attack. The victim, who gave his name as Moshe, told BBC News he was approached while working at a property in Slough on Monday when a man began shouting abuse. Video footage filmed by Moshe shows an aggressive man threatening to break his jaw and attempting to knock a phone from his hand. Thames Valley Police said a 48-year-old man from Slough has been arrested on suspicion of racially or religiously aggravated assault and public order offences. Moshe said he was wearing a black kippa, a traditional head covering worn by Jewish men, at the time of the incident on Elliman Avenue. "I was thinking two things," he said. "Number one was survival. Number two, I was just berating myself: 'What were you thinking going out like this, in England, as a visible Jew?'" In the video, the man can also be seen repeatedly swearing, asking "what are you doing round here, bro?" and calling Moshe a "dirty Jew". He said the incident had left him feeling unsafe in some parts of the country. Police said officers were called at 16:36 BST and that a man in his 20s had been threatened. Det Insp Terry Dixon said the incident would "cause a great deal of concern" within Jewish communities. He added: "Antisemitic incidents, or indeed any hate-related incident, will not be tolerated and we will do everything in our power to bring those responsible to justice." Shomrim, a volunteer group that supports Jewish communities, described the incident as an "abhorrent, vile and unprovoked antisemitic assault".

Apr 23, 2026 · via bbc.com

Enhancing ancient human ruins classification with residual neural networks using visible ...

Abstract Ancient cultural ruins, as tangible evidence of human-environment interaction, provide invaluable historical insights and critical resources for addressing global challenges like climate change and population growth. However, most ruins are indistinguishable and challenging to identify, as they have merged with the surrounding sediments through prolonged natural accumulation and anthropogenic activities. Using SNV in conjunction with the ResNet50 model, this study proposes a novel method for achieving high-precision classification of spectral data from ancient human ruins. The ResNet50 model achieves a classification accuracy of 94.86% when used independently, whereas the accuracy improves to 96.60% when the ResNet50 model is combined with SNV. The model exhibits exceptional classification performance, even when trained with a limited number of spectral image samples from ancient ruins. The superiority of the SNV + ResNet50 model provides a pioneering and effective method for the rapid and accurate identification of ancient human relics using visible-near-infrared spectroscopy. Similar content being viewed by others Introduction Since their emergence, human beings have continuously interacted with the natural environment, accumulating a rich legacy of cultural remains throughout this prolonged and dynamic process. These ruins encapsulate the history of human cognition, adaptation, utilization, and transformation of nature, acting as a repository of extensive information on environmental evolution and human-land interactions. Therefore, ancient ruins can provide invaluable historical insights and serve as crucial resources for addressing global challenges such as climate change and population growth, thereby promoting the sustainable development of human society1,2,3. Under the influence of prolonged natural accumulation and anthropogenic activities, most human ruins have gradually become integrated into the surrounding environment4,5,6. Consequently, differentiating these remains from natural sediments and delineating their spatial morphology and distribution poses a significant challenge. The traditional manual investigation is not only inefficient but also heavily reliant on subjective experience for interpretation7,8,9. It is imperative to

Apr 23, 2026 · via nature.com

ESP-Claw Lets You Build IoT Projects via Chat

ESP-Claw Lets You Build IoT Projects via Chat Using ESP-Claw, you can chat with your ESP32 board to bring hardware projects to life without writing a single line of code. Are you still spending hours writing firmware for your embedded systems projects? If you are using an ESP32 microcontroller, you might not need to anymore — at least in some cases. Espressif has just released ESP-Claw, a chat coding AI agent framework for IoT devices. You simply flash the ESP-Claw firmware to your development board and wire it to any additional hardware you need, then send a message to ESP-Claw via a messaging app like Telegram explaining how it should work. Within seconds, it just works. Users can define device behavior through natural language, with the system dynamically generating Lua scripts to control hardware like LED strips, displays, or motors. Once validated, those scripts can be stored locally, ensuring reliable operation even without a connection to the Large Language Models (LLMs) that created them. This hybrid model — LLM-generated logic combined with deterministic Lua execution — offers both flexibility and stability. ESP-Claw is built around an event-driven architecture. Rather than waiting for explicit user commands, devices can react proactively to real-world triggers such as sensor readings or system events. For time-sensitive applications, local rules execute in milliseconds, bypassing any need for cloud interaction. If no rule applies, the system can escalate to an AI model for deeper analysis or even offload complex tasks like image recognition. Unlike typical AI assistants that lose context between sessions, ESP-Claw organizes long-term data into categories such as user preferences, device events, and behavioral rules. These are stored locally with lightweight tagging, allowing the system to retrieve relevant context efficiently without heavy database overhead. Over time, the device can identify patterns and suggest new automations.

Apr 23, 2026 · via hackster.io

Panasonic adds QR-based biometric onboarding to streamline site access

Panasonic adds QR-based biometric onboarding to streamline site access Panasonic Connect has introduced a new QR‑based face registration feature for its KPAS Cloud access management service. The new feature is aimed at streamlining biometric onboarding at factories, construction sites and other high‑traffic facilities. In 2025 the Japanese company launched the KPAS Cloud Site Management Service. It uses Panasonic’s facial recognition technology to control entry and exit for contractors, staff and visitors. The company says the new QR code feature is designed to remove one of the biggest operational bottlenecks as administrators previously had to collect face images in advance or register users on site. Under the updated workflow, administrators generate a QR code from the management portal and send it to the user by email. On arrival the user scans the QR code at a dedicated terminal and captures their facial image. They then complete registration and gain access to authorized areas. Panasonic says the QR codes contain restricted identification data that can only be interpreted in authorized environments. This prevents third‑party access and the mechanism is now under patent application, the company said in a post on its website (in Japanese and machine translated into English). The update is part of a broader plan to expand KPAS Cloud into a unified platform for on‑site operations, according to Panasonic. Future additions will include facial recognition‑based attendance management, door‑level access control and integration with security camera video feeds. Panasonic says the goal is to consolidate functions that are currently handled by separate systems and provide a more comprehensive digital identity and access management for industrial and logistics environments. From Pakistan to the U.S. and the Philippines, QR codes are linking up with identity verification. In the U.S., CMS is aiming to modernize patient data access with patients able to take a

Apr 23, 2026 · via biometricupdate.com

Interactive Data Experiences : Macrodata Refinement

Macrodata Refinement Lets Users Classify Data Through Emotions Ellen Smith — April 23, 2026 — Tech References: teal-moxie-e7f0e6.netlify.app Macrodata Refinement Application is an interactive digital experience that combines data categorisation with abstract, emotion-based classification. Users engage with a retro-inspired interface to sort numerical data into predefined emotional categories such as Woe, Frolic, Dread, and Malice. The process blends analytical input with subjective interpretation, creating a hybrid task that is both structured and intuitive. Rather than serving a traditional data processing function, the application is designed to simulate or gamify decision-making and pattern recognition. It is typically positioned as an experiential or conceptual tool, appealing to users interested in immersive or narrative-driven interactions. The application reflects broader trends in interactive design where elements of gamification, nostalgia, and abstract systems are combined to create engagement. Its structure highlights how digital interfaces can be used to explore unconventional approaches to data interpretation. Image Credit: Macrodata Refinement The process blends analytical input with subjective interpretation, creating a hybrid task that is both structured and intuitive. Rather than serving a traditional data processing function, the application is designed to simulate or gamify decision-making and pattern recognition. It is typically positioned as an experiential or conceptual tool, appealing to users interested in immersive or narrative-driven interactions. The application reflects broader trends in interactive design where elements of gamification, nostalgia, and abstract systems are combined to create engagement. Its structure highlights how digital interfaces can be used to explore unconventional approaches to data interpretation. Image Credit: Macrodata Refinement Trend Themes - Emotion-based Data Classification — Translating numeric datasets into affective categories creates potential for models that surface sentiment-informed patterns and predictive signals not apparent in traditional metrics. - Gamified Data Interfaces — Interactive, game-like sorting experiences suggest novel engagement frameworks where user-driven play generates labeled datasets and

Apr 23, 2026 · via trendhunter.com

Three arrests in Slough after police live facial recognition use

Three arrests from town centre facial recognition Three people were arrested after a police force used live facial recognition (LFR) vans in a town for a second time, figures show. Thames Valley Police (TVP) used LFR cameras in Slough's High Street on Tuesday and scanned 7,729 faces between about 11:00 BST and 12:45. The technology has been criticised by some civil liberty groups but a challenge against the Metropolitan Police's use of it was dismissed at the High Court the same day. Following the case's conclusion, policing minister Sarah Jones said law-abiding citizens have "nothing to fear" as the technology "only locates specifically wanted people". Youth worker Shaun Thompson and Silkie Carlo, director of campaign group Big Brother Watch, brought the case because of concerns that LFR could be used arbitrarily or in a discriminatory way. Thompson was misidentified as his brother by LFR in 2024 and stopped, detained and questioned by police in London after he was matched by the technology. At the time his brother was on bail for a suspected violent offence. TVP's figures show about 52,600 faces were scanned at the Eden Centre in High Wycombe on 3 March. Though two people were arrested, at least one was not connected to LFR's deployment. It was also used at The Lexicon in Bracknell on 5 March, where just over 20,000 people's faces were scanned. Another three people were arrested after TVP's vans were deployed to Cornmarket Street in Oxford on 20 March. About 37,700 people's faces were scanned there. They were also used in Market Square in Aylesbury on 13 April and in Chesham's High Street on 20 April. About 12,100 and 6,400 people's faces were scanned there, respectively but no one was arrested.

Apr 23, 2026 · via bbc.com

3D foundation model for generalizable disease detection in head computed tomography

Abstract Head computed tomography (CT) imaging is a widely used imaging modality with multitudes of medical indications, particularly in assessing pathology of the brain, skull and cerebrovascular system. It is commonly used as the first-line imaging in neurologic emergencies given its rapidity of image acquisition, safety, cost and ubiquity. Deep learning models may facilitate detection of a wide range of diseases. However, the scarcity of high-quality labels and annotations, particularly among less common conditions, substantially hinders the development of powerful models. To address this challenge, we introduce FM-HCT, a Foundation Model for Head CT for generalizable disease detection, trained using self-supervised learning. Our approach pretrains a deep learning model on a large, diverse dataset of 361,663 non-contrast 3D head CT scans without the need for manual annotations, enabling the model to learn robust, generalizable features. Our results demonstrate that the self-supervised foundation model substantially improves performance on downstream diagnostic tasks compared to models trained from scratch and previous 3D CT foundation models trained on scarce annotated datasets. Similar content being viewed by others Main Head computed tomography (CT) is widely used for rapid evaluation of neurological emergencies such as trauma, haemorrhage and stroke. Although CT is faster, more accessible and less expensive than magnetic resonance imaging (MRI), it provides lower soft-tissue contrast, limiting sensitivity for many neurological conditions. Improving diagnostic capability from CT therefore provides substantial clinical value. Artificial intelligence (AI) has the potential to enhance CT interpretation and support clinical decision-making by enabling earlier and more accurate diagnosis. However, progress in AI-based head CT analysis remains limited by both data availability and model design. Public datasets such as RSNA1 and CQ500 (refs. 2,3) are relatively small and primarily focus on haemorrhage detection, restricting broader clinical applicability. In addition, many existing approaches rely on two-dimensional (2D) convolutional networks that process