Live facial recognition cameras will target knife crime hotspots under new plans to slash rates by a third - Get your news delivered straight to you - sign up to the Morning Mail newsletter for FREE to be first with the day's biggest stories Live facial recognition cameras are to be introduced across the UK in knife crime hotspots under new plans to slash offences by a third. New national mapping technology is being shared with police which is capable of identifying the worst knife crime hotspots down to 100 square metres, even identifying the specific times it is most likely to occur. Forces will be able to switch on cameras to catch out offenders on specific streets associated with knife crime. Under a plan to slash rates by a third in two years, police will also be able to use visible police patrols, CCTV cameras and knife detection arches in an unprecedented crackdown. On Tuesday, ministers announced a £26million Knife Crime Concentrations Fund will be allocated to the 27 police forces in England and Wales that deal with 90 per cent of knife crime. The move comes after live facial recognition camera trials have shown impressive results, with police in the first UK town to get a static camera catching criminals every 34 minutes, spotting suspects on the run for 20 years in some cases. In October, Scotland Yard installed fixed cameras on lamp posts at the entrance and exit to Croydon town centre during a trial which resulted in over 100 arrests for offences including possession of knives, strangulation and sexual assault. The cameras work by taking digital images of passing pedestrians, feeding them into a computer using biometric software to measure facial features. Live facial recognition cameras are to be introduced across the UK in knife crime
Apr 7, 2026 · via dailymail.co.uk
License plate readers have become a familiar part of everyday physical security and access control. What started as a specialized law-enforcement tool is now used on toll roads, in parking garages, at gated entrances, and throughout city infrastructure to quickly identify vehicles at scale. As LPR technology shows up in more places, people are asking tougher questions about how accurate it is, who’s watching over it, and what it means for privacy. This article explains how these systems work, where they are commonly used, what data they collect, and why they are still important in ongoing legal and ethical discussions. Key Takeaways - License plate readers (LPRs) are fast cameras that take pictures of license plates, record the time, and note the GPS location of the reader, creating a timestamped location record of the vehicle - Law enforcement uses LPRs to find stolen cars, locate people with outstanding warrants, and help with criminal investigations. Commercial operators use them for toll collection, parking management, and controlling access to areas. - Collecting a lot of license plate data raises important privacy and civil rights concerns, especially when this data is kept for a long time or shared beyond where it was collected. - LPR systems collect data on billions of vehicles, documenting the movements of individuals who are not connected to any criminal activity. This increases concerns about the surveillance capabilities these systems can provide. - The actual accuracy of LPR systems is often lower than what vendors claim. Their performance can vary greatly depending on environmental factors and conditions. What Is a License Plate Reader (LPR/LPR/ANPR)? A license plate reader is a surveillance technology designed to automatically capture and interpret vehicle registration plates from images or video footage. These systems go by several names: LPR (License Plate Reader), LPR (Automated License
Apr 7, 2026 · via omnilert.com
OCR Studio KYC onboarding tool for APAC recognizes diverse character sets Identity verification provider OCR Studio, which specializes in advanced optical character recognition, is unveiling an AI system for Know Your Customer (KYC) and client onboarding designed for the APAC region. A release from the firm says its APAC-Focused AI System, showcasing this week at GITEX AI Asia in Singapore, addresses the complex challenge of recognizing languages that can involve thousands of character classes, cursive writing, and unusual letter connections. The on-device tool “addresses these scripts’ features and reliably recognizes ID cards, passports, driving licenses, and other identity documents issued across the region.” Its scope covers identity documents in the region’s most widely used logographic, alphabetic, abugida and syllabary writing systems and languages, including Chinese, Japanese, Korean, Thai and Urdu. By keeping the KYC process on device, with no biometric data sent to clouds or external servers, the product helps companies comply with stringent data protection rules in certain APAC nations, such as Singapore’s Personal Data Protection Act (PDPA), China’s Personal Information Protection Law (PIPL) and Indonesia’s Law No. 27 on Personal Data Protection (PDP Law). The announcement follows the launch of OCR’s specialized anti-fraud toolset for advanced on-device document forensics, which includes three layers. The first considers data printing techniques and font controls, format and validity checks, and includes an image tampering detection module that identifies deepfakes, synthetic media and morphed IDs. The second enables document integrity checks and forgery detection, identifying inconsistencies across data fields. And the third uses selfie-to-ID biometric face matching to check whether the legitimate document holder is genuinely present. On April 23, OCR Studios will host a webinar focused on navigating compliance in the EU regulatory landscape. The eIDAS 2.0 regulation underpins the EU’s digital identity infrastructure and ecosystem. The webinar, “eIDAS 2.0 and
Apr 7, 2026 · via biometricupdate.com
The Ouroboros of Faces - Dates2025 - Ongoing - Author - Locations Berlin, Helsinki, Hong Kong - Shortlisted The advent of synthetic face data marks the latest phase of facial recognition. First, humans taught machines how to see faces. Now, machines are trained to recognize faces made by other machines —the logic forms a self-devouring snake, an ouroboros. Synthetic face data are deepfakes or digitally rendered 3D faces generated solely as training material for facial recognition algorithms. Technologists have proposed them as a solution to the insatiable data demands of large-scale AI models, claiming the added benefit of circumventing issues of privacy and consent. These statistical artefacts of facial recognition are fed back into the models themselves. The advent of synthetic face data marks the latest phase of facial recognition, where the logic of statistical inference is stretched to its breaking point. First, humans taught machines how to see faces. Then, networks learned to generate digital faces modeled on human ones. Now, machines are trained to recognize faces made by other machines—a self-devouring snake, an ouroboros. In the near future, our faces may be judged by algorithms that have never “seen” a real one. Through the machinic eye, a facial recognition algorithm looks back at us. They claim the power to predict sexuality and political ideology from facial landmarks. Like a magic mirror, it was believed to have the sacred power of statistical prophecy. Sacred powers are rarely questioned. When facial recognition fails, the error is attributed to flawed algorithms or noisy training data, but rarely the underlying logic of physiognomy itself: the assumption that a person can be known, measured, and judged by appearance. This project expands on my ongoing project (Inter)Faces of Predictions, which offers a critical reading of facial recognition alongside the forgotten pseudo-science of physiognomy and
Apr 7, 2026 · via phmuseum.com
Abstract Clouds significantly affects the power output of solar energy systems, and it also decrease the life of modules in photovoltaic system and receivers in the concentrating solar power system, resulting in costs of operation and maintenance. In the present study, the MRR-YOLO model was proposed, which was based on deep learning and instance segmentation technique. The MSDA, the RCS-OSA, and the RFAConv modules were used, and their functions to the cloud segmentation were investigated. Results found that the MSDA module helped maintain the model’s lightweight, the RFAConv module had a better feature extraction of the clouds. The instance segmentation method was better than semantic segmentation in clouds detection, especially for clouds of varying shapes. The PB, the RB, and the mAP50B of the MRR-YOLO model were 79.2%, 66%, and 74.7%, respectively. For the segmentation task, the PM, the RM, and the mAP50M were 79.3%, 64.8%, and 73%, respectively. The MRR-YOLO model was validated by the SWIMSEG, the CCSN, the all-sky datasets, and the real cloud images in Zhengzhou city, it had a better detection performance and applicability than other models. A heatmap comparison was also conducted, showing that the model accurately detected all features of the clouds. Similar content being viewed by others Data availability The datasets used and/or analysed during the current study available from the corresponding author on reasonable request. References Abdalla, S. A mathematical model for economic and prognostic studies of solar photovoltaic power: Application to China, the EU, the USA, Japan and India compared to worldwide production. Renew. Energy Focus. 50, 100607 (2024). Borodiņecs, A., Lebedeva, K. & Odineca, T. Evaluation of household electricity consumption in multi-apartment buildings for optimization of rooftop PV systems. Energy Build., 114971. (2024). Rakhmatov, A. et al. Advancements in renewable energy sources (solar and geothermal): A brief review[C]//E3S Web of
Apr 7, 2026 · via nature.com
Fujisoftis developing an AI-enhanced physical security system built on the AMD Embedded+ platform that integrates FPGA adaptability with x86 processing performance into a unified architecture.
Fujisoft’s next-generation security system moves beyond merely detecting motion. It enables AI-based image recognition that dramatically increases detection accuracy and reduces false positives. Fujisoft completed a demonstration unit in 2025 and is now refining the solution for broader deployment.
At the core of the solution is the AMD Embedded+ platform. It combines an AMD Ryzen™ Embedded processor and a Versal™ adaptive SoC on a single board to enable AI inference and low-latency processing on a variety of sensors using programmable I/O. The pre-integrated architecture minimizes design complexity and costs, and it helps accelerate time to market.
“By combining both adaptive and x86 embedded technologies into a single, integrated solution, the AMD Embedded+ platform fully leverages Fujisoft’s strengths in both hardware and software development,” said Naoya Yanagitsubo, Fujisoft manager. “Through conversations with various customers, we have identified specific challenges, and we believe that Embedded+ is well positioned to address and solve these issues.”
𝐒𝐭𝐚𝐲 𝐢𝐧𝐟𝐨𝐫𝐦𝐞𝐝 𝐰𝐢𝐭𝐡 𝐨𝐮𝐫 𝐥𝐚𝐭𝐞𝐬𝐭 𝐮𝐩𝐝𝐚𝐭𝐞𝐬 𝐛𝐲 𝐣𝐨𝐢𝐧𝐢𝐧𝐠 𝐭𝐡𝐞 WhatsApp Channel now! 👈📲
𝑭𝒐𝒍𝒍𝒐𝒘 𝑶𝒖𝒓 𝑺𝒐𝒄𝒊𝒂𝒍 𝑴𝒆𝒅𝒊𝒂 𝑷𝒂𝒈𝒆𝐬 👉 Facebook, LinkedIn, Twitter, Instagram
Apr 7, 2026 · via digitalterminal.in
Over-the-Air Computation Uses Radio Interference to Crunch Data Sensor networks could improve capacity faster than they grow Picture a highway with networked autonomous cars driving along it. On a serene, cloudless day, these cars need only exchange thimblefuls of data with one another. Now picture the same stretch in a sudden snow squall: The cars rapidly need to share vast amounts of essential new data about slippery roads, emergency braking, and changing conditions. These two very different scenarios involve vehicle networks with very different computational loads. Eavesdropping on network traffic using a ham radio, you wouldn’t hear much static on the line on a clear, calm day. On the other hand, sudden whiteout conditions on a wintry day would sound like a cacophony of sensor readings and network chatter. Normally this cacophony would mean two simultaneous problems: congested communications and a rising demand for computing power to handle all the data. But what if the network itself could expand its processing capabilities with every rising decibel of chatter and with every sensor’s chirp? Traditional wireless networks treat communication as separate from computation. First you move data, then you process it. However, an emerging new paradigm called over-the-air computation (OAC) could fundamentally change the game. First proposed in 2005 and recently developed and prototyped by a number of teams around the world, including ours, OAC combines communication and computation into a single framework. This means that an OAC sensor network—whether shared among autonomous vehicles, Internet-of-Things sensors, smart-home devices, or smart-city infrastructure—can carry some of the network’s computing burden as conditions demand. The idea takes advantage of a basic physical fact of electromagnetic radiation: When multiple devices transmit simultaneously, their wireless signals naturally combine in the air. Normally, such cross talk is seen as interference, which radios are designed to suppress—especially digital
Apr 7, 2026 · via spectrum.ieee.org
Abstract Identifying changes in satellite images is vital for tasks like tracking land cover and land use, evaluating disaster impacts, and conducting military surveillance. Although conventional techniques for detecting changes in multispectral remote sensing data are commonly applied, they often fail to meet the requirements for reliability and precision. Recently, deep learning methods have emerged, providing more accurate and effective solutions for monitoring environmental transformations and urban expansion in satellite imagery. This paper introduces TransSiamUNet, a deep learning architecture that combines Siamese networks, U-Net segmentation, and Vision Transformers (ViT) for high-precision change detection. The model processes paired Sentinel-2 images via a tailored preprocessing pipeline and integrates local and global feature extraction for pixel-level change segmentation. On the OSCD benchmark, TransSiamUNet achieves an accuracy of 0.94, surpassing the Siamese network (0.86), U-Net (0.84), and Siamese+U-Net hybrid (0.91). These results demonstrate the model’s superior capability in detecting fine-grained urban and environmental changes, highlighting its suitability for real-world remote sensing applications. Data availability Data and code used for training, evaluation, and experimentation in this study are available on request. Code availability The code used for training, evaluation, and experimentation in this study is available on request. Materials availability All datasets and code used in this study are available. The Arabic summarization dataset and model implementations can be available on request. References Richards, J. A. Remote sensing digital image analysis: An introduction (Springer, Cham, 2023a). Liu, X., Zhang, H. & Chen, Y. Urban expansion detection using ndbi and remote sensing data: A case study of rapid urbanization. Remote Sensing 15(4), 1123–1138 (2023a). Hughes, L. H., Schmitt, M., Mou, L., Wang, Y. & Zhu, X. X. Identifying corresponding patches in sar and optical images with a pseudo-siamese cnn. IEEE Geoscience and Remote Sensing Letters 15(5), 784–788 (2018). Daudt, R.C., Le Saux, B. & Boulch, A.
Apr 7, 2026 · via nature.com
Abstract Deep learning has demonstrated remarkable success in augmenting fluorescence imaging under photon-limited conditions. However, existing restoration networks are typically devised for training with augmented patches far smaller than the full-view raw data, an overlooked aspect that compromises fidelity and noise-resistance due to the loss of global statistics. To address this limitation, we propose a large-patch network (LargePNet), which synergizes the large effective receptive field provided by shallow ultra-large-kernel convolutions and the nonlinear representation capabilities of deep networks through scale separation. It effectively and efficiently leverages large-view global information for restoration. Directly trained with large-view images, LargePNet shows contrasting advantages over state-of-the-art small-patch networks, with 0.5-2 dB higher peak signal-to-noise ratio across eight representative restoration tasks, involving implementations for single-image, video, and volumetric fluorescence data. For full-view processing, LargePNet generally holds around 4-fold and 20-fold higher computational efficiency compared to advanced convolution-based and Transformer-based networks, respectively. The assistance of LargePNet helps achieve 30-hour-long fluorescence imaging to monitor cytoskeleton dynamics, and hour-long tri-color super-resolution imaging to investigate organelle interaction, showcasing its advancement in live-cell imaging. Similar content being viewed by others Data availability The open-source data used in this study are all publicly available, as listed in Supplementary Table 15. The self-established dataset of STED denoising/deblurring, virtual sampling of SMLM, volumetric background removal datasets, and the source training data for the open-source BioSR and BioTISR datasets are available: https://zenodo.org/records/15694668. Code availability The source Python code of the LargePNet series, including LargePNet (for single-image restoration), LargeP-GAN (for generative single-image restoration), LargeP-TISR (for time-lapse video restoration), and 3D-LargePNet (for volumetric data restoration) are all publicly available at the GitHub repository56: https://github.com/YiweiHou/LargePNet-for-fluorescence-image-restoration. Trained LargePNet models that can reproduce the results in the paper are available at: https://figshare.com/s/05f576c96b08add7eee0. References Ledig, C. et al. Photo-realistic single image super-resolution using a generative adversarial network. In Proc. IEEE
Apr 7, 2026 · via nature.com
Automates critical DODA processes from document extraction to SAT synchronization for customs agencies, freight forwarders and cross-border carriers navigating Mexico’s 2026 Customs Law Reform. HOUSTON – April 7, 2026 – Freight Technologies, Inc. (Nasdaq: FRGT, “Fr8Tech” or the “Company”), a technology-centric logistics company offering a diversified portfolio of AI software solutions designed to address key inefficiencies in the supply chain, today announced the commercial launch of DODA Smart, an AI-powered platform to automatically verify, monitor, and trace the full lifecycle management of Digital Customs Documents (“DODAs”) for customs agencies, freight forwarders, carriers, and import-exporters operating across the U.S.–México corridor. The platform is designed to synchronize automatically with the México Tax Administration Service (Servicio de Administración Tributaria, or “SAT”) to provide real-time compliance intelligence. Developed by Fr8Tech’s in-house AI Lab, DODA Smart is designed to help customers address new digital recordkeeping and traceability requirements associated with Mexico’s amended 2026 Customs Law Reform. DODA Smart is designed to reduce reliance on physical paper records, manual QR-code reading, fragmented document tracking, and reactive compliance workflows. It is designed to provide automated SAT synchronization and AI-assisted document verification for participants across the customs clearance chain — from the customs agency to the freight forwarder to the carrier. DODA Smart is a natural extension of Fr8Tech’s cross-border intelligence ecosystem and fully integrates with the Company’s existing suite of solutions, completing an essential compliance layer for freight operations across the U.S.–Mexico border. The ability to manage audit-ready customs documentation at scale — quickly, efficiently and accurately — is more important than ever, as cross-border volume grows amid increased nearshoring activity. DODA Smart is a strategic differentiator for any operator moving goods across the border. Javier Selgas, Chief Executive Officer of Freight Technologies, said, “DODA Smart is the next logical step in Fr8Tech’s mission to remove friction
Apr 7, 2026 · via fr8technologies.com
By Ayathandwa Tsili When artificial intelligence (AI) recognises a face, reads a medical scan or makes decisions, it often does so in ways that even its creators say they cannot fully explain. But for Rhodes University MSc Applied Mathematics graduate Georgina Fiorentinos, that wasn’t good enough – so she set out to find out for herself. And she wanted more than a mathematical answer. She wanted to understand how machines interpret the world around them and how reliable those interpretations really are. Her MSc research explored the inner workings of deep learning models, the type of AI behind technologies such as image recognition systems and automated decision-making tools. While these systems are widely used, the processes inside them are often difficult to interpret. “I’ve always been interested in what is happening inside these models,” Fiorentinos explains. “We usually see the input and the output, but the actual process in between can feel like a black box. I wanted to understand how the model decides what information is important and what it ignores.” Her interest began during her Honours research, where she studied how compression techniques could reduce noise in deep convolutional neural networks, models commonly used to analyse images. That work led her to investigate how information changes as it moves through a neural network and whether important details are preserved along the way. Working within the Rhodes University Artificial Intelligence Research Group, and supervised by Professor Atemkeng, she focused on the relationship between ‘signal’ and ‘noise’ inside these systems. In simple terms, signal refers to useful information a model needs to make accurate predictions, while noise refers to irrelevant details that can interfere with those decisions. “An image might seem obvious to us, but to a computer it’s just numbers,” she says. “The model has to learn which patterns
Apr 7, 2026 · via ru.ac.za
This morning, April 7, 2026 (early morning JST), I have released the prototype of UNDEFINED: ILS for free on itch.io — a minimalist puzzle adventure where a local LLM evaluates how players respond to visual incompleteness and contradictions. Players are shown "missing," "contradictory," or "unfinished" visual structures and respond by drawing lines directly on the screen. The AI immediately judges the response with "PROBLEM WAS SOLVED" or "PROBLEM WAS NOT SOLVED" and explains its reasoning in natural language. The way players draw, retry, or skip accumulates as behavioral patterns that influence the game's outcome. This type of interactive experience only became feasible in 2026 due to significant improvements in local LLM image recognition. The prototype includes 10 selected problems out of 100 (ending not yet implemented). I'm releasing it for free to gather wide feedback ahead of the full Steam version. # Key Features Minimalist input: Draw one line — complete gaps, resolve contradictions, redefine rules, or reject the problem. Real-time AI judgment with natural language explanations — it feels like your thinking is being calmly observed. No hints, timers, or limits — the challenge lies in choosing your "attitude" toward ambiguity. Player behavior patterns (not just right/wrong answers) shape the narrative. # Example Problems Complete the unfinished equation "1 + 1 =" or leave it as is? Fill a missing face, highlight the gap, or reinterpret the pattern entirely? In a blocked road scenario, draw a path, change the rules, or reject the premise? # Specs Title : UNDEFINED: ILS Genre : Puzzle Adventure / Single Player Planned Release Date : Autumn 2026 Languages : English, Japanese, Chinese（Simplified / Traditional), Korean Download the Free Prototype (Available Now) : https://kamekichi.itch.io/undefined-ils Steam Store Page : https://store.steampowered.com/app/4481900 Play Demo Movie : https://youtu.be/VQGQtjyYXv4 System Requirements : Windows 10/11 – VRAM 6GB+ GPU recommended
Apr 7, 2026 · via gamespress.com
Tokyo – April 7, 2026 – TOPPAN Holdings Inc. (TYO: 7911) (TOPPAN Holdings) and subsidiary TOPPAN Inc. (TOPPAN) have developed an AI-Optical Character Recognition (OCR) engine that can decipher medieval Greek, a type of writing that is generally difficult to understand. Moving forward, thanks to a collaborative relationship with the Vatican Apostolic Library (Vatican Library), the Printing Museum, Tokyo (operated by TOPPAN Holdings) plans to scan images and digitized text from Greek manuscripts belonging to the Vatican Library. These will be used as learning data to refine accuracy, with the goal of increasing TOPPAN’s AI-OCR recognition accuracy to more than 95%. Visuals showing the findings of this approach will be displayed during the “How Masterpieces Are Born: Biblioteca Apostolica Vaticana â ¢+” special exhibition running from April 25, 2026, at the Printing Museum, Tokyo. TOPPANâs AI-OCR Engine Development Old texts hold a myriad of information regarding important historical facts and regional culture. However, many such texts are handwritten and are difficult to read for most people in the modern world. Correctly interpreting these texts to preserve culture is now an important challenge not only in Japan, but also on the global stage. For the past 30 years or so, the TOPPAN Group has been collaborating with the Vatican Library on several projects to safeguard culture for future generations. During that time the Vatican Library has released over two million items from part of its collection to the public for research and educational purposes in the form of IIIF1 high-resolution images. A total of nine million images were released, and that number continues to grow. Additional information such as transcriptions2 and annotations has been added to some of the Greek manuscripts, but experts in medieval Greek would need to work for a long period of time to complete the entire collection.
Apr 7, 2026 · via holdings.toppan.com
Chinese spy cameras remain in widespread use in the United States, in Canada and Mexico and worldwide. They are sold as security and traffic cameras, but that is a cover story for their real use. Such cameras are used for social and political control turning countries into digital prisons. The US and Israel have been able to hack them. In places where they are banned, China is now using drones to conduct spying in place of fixed installation cameras. The recent persistent Barksdale Air Base drone incursion along with many others may be China’s partial answer to the ban on its security cameras. In fact, the rise of drone security incidents corresponds almost exactly to the US 2019 ban on Chinese cameras at American military bases. What is happening in the United States is taking place elsewhere. Back in 2018 I discovered a sole source contract for Hikvision cameras for the US Embassy in Kabul, Afghanistan. Hikvision is Hangzhou Hikvision Digital Technology Co., Ltd. It is a publicly traded company, but it is effectively controlled by the Chinese government. While it is listed on the Shenzhen Stock Exchange (SZSE: 002415), its ownership structure is a mix of state-owned enterprises (SOEs), private founders, and institutional investors where state-owned enterprises control more than half the company. Why, I wondered, was the US Embassy in Kabul ordering Chinese cameras for its security, and why was the contract sole-source, meaning no other camera would qualify for the job? As I began asking questions about the announced contract (a sole source announcement means the deal is done), I found out that Hikvision, along with another Chinese camera company, Dahua Technology (Zhejiang Dahua Technology Co., Ltd.), apparently had weak security amounting to a functional back door. These cameras work through the Internet and can be easily
Apr 7, 2026 · via asiatimes.com
Netflix is making a bigger play for young viewers with the launch of a games app for kids, the renewal of two preschool shows and a new series based on one of the most enduring nursery rhymes of all time. The streamer on Monday launched Netflix Playground, a gaming app aimed at kids 8 and under that features activities with characters from Peppa Pig, Sesame Street, Storybots and other kids’ shows in Netflix’s library. The app is free for all subscribers and promises no in-game purchases or other fees and no advertising (regardless of whether the linked account subscribes to Netflix’s ad-supported tier). Related Stories Games will also be playable offline, should parents not want to burn their wireless hotspot on a long car ride. The Playground app launched Monday in the United States, Canada, U.K., Australia, the Philippines and New Zealand and will debut in the rest of the world on April 28. Along with Playground’s launch, Netflix on Monday ordered an animated preschool series called Young MacDonald. Created by Gabrielle Meyer (Ada Twist, Scientist), the show centers on Mac, the grandson of Old MacDonald who cares for his farm animals and crops with the help of a pig named Dumpling. The streamer has also ordered new seasons of animated shows Trash Truck and The Creature Cases. “We’re building a world where kids can not only watch their favorite stories, they can step inside them and interact with their favorite characters,” said John Derderian, vp animation series and kids & family TV at Netflix. “We’re creating a seamless destination for discovery, learning, and play. Whether it’s reuniting with Hank and the Trash Truck crew for new adventures or making a smoothie with Peppa Pig, watching and playing on Netflix can be the fun and easiest part of every family’s
Apr 6, 2026 · via hollywoodreporter.com
Netflix has released a new iPhone and iPad app called Netflix Playground. Similar to Apple Arcade, the new Netflix Playground app includes ad-free games with no in-app purchase content. Netflix Playground includes 7 ‘mini games’ for younger kids Netflix Playground includes multiple “mini games” that younger kids will enjoy. These include content from Peppa Pig, Sesame Street, Dr. Seuss, and more. At launch, Netflix Playground includes seven mini games inside the new app: - Dr. Seuss’s Red Fish, Blue Fish: Dive into musical play zones filled with rhythmic fun and gentle surprises. - Dr. Seuss’s Horton!: Welcome to the jungle! Horton’s got fun and games that encourage creativity through interactive cause-and-effect play. - Bad Dinosaurs: Prehistoric puzzles, sticker scenes, music, and memory games abound with Janet and her brood of tiny thunderlizards. - Sesame Street: It’s always a sunny day on Sesame Street, and the activities here encourage pattern recognition, object recognition, and free-form play. - Dr. Seuss’s The Sneetches: Your kids will join Stella Sneetch on an imaginative adventure through Starbelly and Moonbelly Villages via creative play activities like the Shape Printer and Vehicle Builder, which encourage imaginative design and pattern recognition. - Peppa Pig: Play, count, care for guinea pigs, decorate cakes, and a whole lot more with Peppa and the rest of her Peppatown crew. - StoryBots: These inquisitive creatures are here to help your kids hone their memory, improve pattern recognition, develop budding musical skills, and more. More titles will be added throughout the year, Netflix says, including content from “Gabby’s Dollhouse, KPop Demon Hunters, PJ Masks, My Little Pony, and PAW Patrol.” The company says Netflix Playground access is part of every membership, not just ad-free and premium tiers. Each mini game can be downloaded individually within Netflix Playground. Once downloaded, mini games are available
Apr 6, 2026 · via 9to5mac.com
Abstract Road crack detection is crucial for road maintenance and traffic safety. To address the low efficiency and limited generalization capability of traditional crack detection methods, this research proposes an improved model dubbed EVA-YOLOv8 (Efficient ViT Attention-YOLOv8) based on the YOLOv8n framework. This model integrates MobileViT Block and an Efficient Multi-scale Attention (EMA) mechanism, while employing the more efficient GhostConv to provide lightweight optimization for the network. Comparisons and ablation experiments were conducted on road crack datasets, and the attention mechanism of the model on crack features was analyzed using Grad CAM + + heatmap. Results indicate: (1) The EVA-YOLOv8 model achieved values of 0.897 (mAP@0.5), 0.706 (mAP@0.5:0.95), 0.907 (Precision), 0.881 (Recall), and 0.894 (F1-score), exhibiting good generalization ability across multiple categories, different scales, and low-contrast scenarios. (2) Compared with the original YOLOv8, the improved model demonstrates enhanced performance: mAP@0.5 and Recall are increased by 6.6% and 5.3%, respectively, while the parameters are reduced by 16.3%. (3) Grad-CAM + + heatmaps indicate that the model not only achieves high response values in the main region of the target but also performs well in target boundary localization and background suppression. This research may provide a reference for related fields. Data availability Data are available from the corresponding author upon reasonable request. References Tian, Z., Shao, X. & Bai, Y. Graph-MambaRoadDet: A Symmetry-Aware Dynamic Graph Framework for Road Damage Detection. Symmetry 17, 1654 (2025). Ma, L. & Chen, M. Road damage detection based on improved YOLO algorithm. Sci. Rep. 15, 28506 (2025). Wang, J. et al. Road defect detection based on improved YOLOv8s model. Sci. Rep. 14, 16758 (2024). Xing, Y. et al. EMG-YOLO: road crack detection algorithm for edge computing devices. Front. Neurorobotics. 18, 181423738–181423738 (2024). Cano-Ortiz, S. et al. Leveraging a deep learning generative model to enhance recognition of minor
Apr 6, 2026 · via nature.com
On 1 June 2025, China’s Cyberspace Administration (CAC) brought into force the Measures for the Security Management of Face Recognition Technology Applications (the „Measures“). This landmark regulation is the first piece of dedicated legislation in China governing the use of biometric facial data, and it carries significant implications for any organization processing face recognition data within Chinese territory. Why Now? China already had a general legal framework in place. The Personal Information Protection Law (PIPL), the Data Security Law, the Cybersecurity Law, and the Network Data Security Regulations all touch on how personal and sensitive data should be handled. The Measures build directly on these, filling a gap with rules specifically designed for face recognition. Face recognition technology has quietly become part of everyday life in China. Airports and train stations use it for identity checks. Residential compounds and office buildings rely on it for access control. Banks and payment platforms deploy it to authenticate transactions. Public security agencies use it to track and identify suspects. All of this has brought real convenience. But it has also raised serious concerns. Unlike a password, facial data cannot be reset once it is leaked. The growing risk of identity fraud, unauthorized profiling, and covert surveillance eventually pushed regulators to act. Who Does This Apply To? According to Article 2 of the Measures, the regulation covers any entity or individual applying face recognition technology to process facial information within mainland China. There is one notable exception. Organizations doing research and development or algorithm training with facial data are not covered by the Measures, though they still have to comply with PIPL and other relevant laws. That said, once the technology moves from the lab into a real-world application, full compliance obligations apply. The Core Compliance Requirements - Mandatory Disclosure (Article 5): Before collecting
Apr 6, 2026 · via datenschutz-notizen.de
Abstract Automated skin lesion detection is essential for early dermatological diagnosis. Most deep learning algorithms employ dermoscopic images and ignore the clinical context of dermatologists. This constraint decreases the robustness and interpretability, particularly in visually ambiguous instances. An explainable multimodal cross attention framework that merges clinical text with dermoscopic images to increase diagnostic accuracy and semantic grounding in automated skin lesion detection. A Text-Guided Cross-Attentive Visual Feature Network multimodal learning architecture (TG-CAVNet) using Bio-ClinicalBERT based clinical text encoding and EfficientNet-B4 visual feature extraction is proposed. The framework uses text-guided channel-wise feature modulation, text-queried cross-attention for semantic-spatial alignment and adaptive multi-stream fusion to merge complementary representations. The model was trained end-to-end using hybrid focal and cross-entropy losses. On a multimodal dermoscopic dataset of 6194 aligned image-text samples, TG-CAVNet outperforms state-of-the-art multimodal baselines with 90.75% accuracy and a macro Jaccard score of 0.82. Ablation investigations are performed that confirmed the separate and synergistic impacts of the components, whereas attention visualizations improved interpretability. Text-guided cross-attentive multimodal learning improves the performance and explainability of automated skin lesion identification. The robust and clinically interpretable decision-support framework TG-CAVNet demonstrates the necessity of integrating semantic clinical context with visual analysis in dermatological AI systems. Similar content being viewed by others Data availability Dataset is available at the link. https://ieee-dataport.org/documents/multimodal-dermoscopic-dataset. References Debelee, T. G. Skin lesion classification and detection using machine learning techniques: A systematic review. Diagnostics. 13(19), 3147. https://doi.org/10.3390/diagnostics13193147 (2023). Naqvi, M., Gilani, S. Q., Syed, T., Marques, O. & Kim, H. C. Skin cancer detection using deep learning—A review. Diagnostics. 13(11), 1911. https://doi.org/10.3390/diagnostics13111911 (2023). Vieira, J., Mendonça, F. & Morgado-Dias, F. Deep learning approaches for skin lesion detection. Electron 14, 2785. https://doi.org/10.3390/electronics14142785 (2025). Haggenmüller, S. et al. Skin cancer classification via convolutional neural networks: systematic review of studies involving human experts. Eur. J. Cancer. 156, 202–216.
Apr 6, 2026 · via nature.com
Facial recognition drones to be deployed in the Channel to catch migrant smugglers steering small boats Facial recognition drones are to be deployed in the Channel to catch migrant smugglers driving the small boats. Military-grade cameras that can identify faces up to half a mile away are now being trialed by Border Force. The radar-enabled technology will be able to spot targets even if they have altered their outfit or appearance. Officers have reported that those piloting the boats change their clothes when they see British rescue vessels to blend in with passengers, or even shave at sea to make them more difficult to recognise. Thus far, this has allowed many to avoid being caught. The Home Office hopes the new strategy, first reported by The Times, will increase the number of prosecutions. Smugglers can be convicted of assisting unlawful immigration - an offence that can carry a sentence of life imprisonment. But the technology must undergo rigorous testing and review in order to overcome practical and legal hurdles before being rolled out. It comes after Shabana Mahmood signed a two-month extension deal for French cops to stop small boats Dangerous crossings in the Channel have increased over the past three years, with 41,472 people arriving in the UK by small boat in 2025 The facial-recognition drones are the latest initiative from the Home Office in its battle to drive down the number of crossings. So far this year, 4,766 migrants have made the journey across the channel, which is down 28 percent on the same period last year. Last week, two people died and three were injured when hundreds of migrants tried to pile onto a rubber dinghy bound for Dover from France's shore. Smugglers took advantage of the improved weather in northern France to launch at least five
Apr 6, 2026 · via dailymail.co.uk