Abstract Stampede incidents in densely populated environments remain a critical challenge for public safety, resulting in severe casualties and significant disruption. This paper presents a novel, data-driven framework for automated stampede detection and crowd risk classification, leveraging the integration of Farneback optical-flow computation with a hybrid CNN-LSTM architecture. Two complementary datasets, UCSD Anomaly detection dataset and Agoraset dataset, were combined to capture a broad spectrum of crowd behaviors and densities. The proposed system classifies crowd states into four distinct risk levels: normal, moderate, dense, and risky, thereby offering a more granular assessment than traditional binary models. The model was trained and evaluated on a balanced dataset of 10,000 annotated frames, with rigorous preprocessing and augmentation to ensure robustness. Experimental results demonstrate an accuracy of 99.75%, further cross-dataset evaluation on the UMN benchmark assesses the robustness under domain shift conditions. While the approach shows strong potential for real-time deployment in public event management and emergency response, current limitations include computational latency and challenges in ultra-dense, occluded scenarios. Similar content being viewed by others Data Availability The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. References Cob-Parro, A.C., Losada-Gutiérrez, C., Marrón-Romera, M., Gardel-Vicente, A., Bravo-Muñoz, I. & Sarker, M.I. A proposal on stampede detection in real environments. In: Proceedings of the IPIN 2021 WiP, Lloret de Mar, Spain (2021). https://ceur-ws.org/Vol-3097/paper29.pdf Sharif, M. H. & Djeraba, C. An entropy approach for abnormal activities detection in video streams. Pattern Recogn. 45(7), 2543–2561. https://doi.org/10.1016/j.patcog.2011.11.023 (2012). Duives, D. C., Oijen, T. & Hoogendoorn, S. P. Enhancing crowd monitoring system functionality through data fusion. Sensors 20(20), 6032. https://doi.org/10.3390/s20216032 (2020). Lalit, R. & Purwar, R. Crowd abnormality detection using optical flow and glcm-based texture features. J. Inf. Techn. Res. 15 (2022) https://doi.org/10.4018/JITR.2022010110 Jadhav, C., Ramteke, R. & Somkunwar, R.
Apr 10, 2026 · via nature.com
Distracted driver and seatbelt detection cameras target drivers who use portable devices while driving. They also can pick up drivers and front seat passengers not wearing their seatbelt. Illegally using a portable device, such as a mobile phone, when driving significantly increases the risk of being involved in a serious crash. In Victoria, widespread roll-out of automated mobile phone enforcement is predicted to prevent 95 casualty crashes per year. What is a portable device? A portable device is any electronic device that can communicate wirelessly and display information but is not an inbuilt device, mounted device, motor bike helmet device or wearable device. Examples of portable devices include: - a dispatch system - a tablet or iPad - a mobile phone - a media player - a camera - a laptop - an information, navigation or entertainment system - a video game console. CB radios and other two-way radios are not included. For full definitions see Road Safety Road Rules 2017. How the cameras work Each camera trailer has 2 cameras and an infra-red flash that are enabled with artificial intelligence (AI) software. The cameras take high-resolution images any time of the day or night, and in all traffic and weather conditions. The AI technology automatically reviews each image. If it does not see a potential offence, it will reject the image. If the AI detects a driver who may be using a portable device or not wearing their seatbelt it flags the image for further review. Images where there might be a potential offence are then checked and verified by qualified independent officers. Illegally using a mobile phone when driving significantly increases the risk of being involved in a serious crash. Images captured by the cameras If you receive an infringement, you can view the images captured at fines.vic.gov.au.
Apr 10, 2026 · via vic.gov.au
Abstract Group activity recognition requires a holistic understanding of individual actions, their spatial relationships, and the surrounding environment. Traditional methods that focus solely on isolated movements often fail to capture the complex inter-player and scene-level dependencies inherent in sports and crowd scenarios. In this research work, a model for group activity recognition is developed. The proposed model combines various contextual features through the integration of poses of individual actors in the scene with the pose-aligned spatial scene context for relational reasoning. Pose features of individual actors are extracted using mmPose, while the scene-level context is encoded through pose-conditioned spatial feature aggregation rather than explicit semantic segmentation. These pose and scene context features extracted are combined and used to construct Actor Relation Graphs (ARGs) using Zero Normalized Cross Correlation (ZNCC) which improves robustness to appearance and variations in illumination. Further, Graph Convolutional Networks (GCNs) are modelled using relationships between individual actors in a scene and their group activities. The proposed framework explicitly combines pose-level and scene-level contextual features into a single relational graph, in contrast to previous ARG-GCN approaches that mainly rely on appearance features. The model is evaluated on two benchmark datasets: the Collective Activity dataset (CAD) and the Volleyball dataset (VD). The model exhibits classification accuracies of 95.02% and 94.81% on CAD and VD, respectively. On a TITAN-XP GPU, the average time per video clip with 41 frames is approximately 0.2 s. The results show that the combination of pose and scene contexts features enhances graph-based relational learning and improves recognition accuracy. Similar content being viewed by others Data availability The Volleyball dataset is publicly available in the “mostafa-saad/deep-activity-rec **”** repository, [https://github.com/mostafa-saad/deep-activity-rec? tab=readme-ov-file#dataset](https:/github.com/mostafa-saad/deep-activity-rec? tab=readme-ov-file) and Collective Activity Dataset is publicly available in the [Computational Vision and Geometry Lab (CVGL) website at Stanford University](https:/cvgl.stanford.edu/projects/collective/collectiveActivity.html) , [https://cvgl.stanford.edu/projects/collective/collectiveActivity.html](https:/cvgl.stanford.edu/projects/collective/collectiveActivity.html) . The datasets used
Apr 10, 2026 · via nature.com
HYDERABAD: The Telangana Higher Education department has planned to make a facial recognition-based attendance system mandatory in all government and private degree colleges from the upcoming academic year to ensure accurate attendance tracking and enhance student welfare. The decision was taken during a project monitoring unit (PMU) meeting chaired by Education Secretary Yogita Rana on Thursday to review key aspects of admissions, attendance and data management in higher education institutions across the state. One of the key decisions taken at the meeting was that the Degree Online Services Telangana (DOST) notification for the 2026-27 academic year will be issued only after the announcement of Intermediate results. In a move to strengthen monitoring, the meeting made it mandatory for both government and private colleges to implement a facial recognition-based attendance system, aimed at ensuring accurate tracking and improving student welfare. The PMU also directed all institutions to strictly enforce mandatory attendance. Universities have been asked to issue circulars to affiliated colleges to ensure compliance, with a focus on improving academic discipline and learning outcomes. During the meeting, TGCHE chairman Balakista Reddy highlighted the introduction of new and innovative undergraduate courses from the 2026-27 academic year, stating that extensive consultations were held with vice-chancellors, subject experts and other stakeholders. He said colleges will be required to enter into MoUs for student internships, while syllabus structures will be provided by universities, along with AI-enabled, tutor-supported learning materials to be made available online. Addressing admission-related concerns, the meeting resolved that students in colleges where DOST admissions are 15% or less will be allowed to opt for sliding to other institutions through the platform to secure better opportunities.
Apr 10, 2026 · via newindianexpress.com
Street photography was built on proximity, on the unscripted moment when two strangers briefly shared the same space and the same gaze. In a world where every face is searchable, traceable, and legally accountable, that proximity no longer carries the same meaning. The future of street photography doesn't look like a sudden collapse. It looks like a slow, managed retreat. On the surface, the genre seems healthy: cameras are easier to use, and cities are more crowded than ever. But the shift isn't about aesthetic trends. It's about a fundamental change in the status of the human face. Street photography was built on the "friction of the second" — the unscripted encounter between strangers. In this tradition, the face wasn't an accessory; it was the raw material of the craft. Now, that material has become toxic. The idea of the "random passerby" is eroding as a category. In a data-driven environment, the person in the frame is no longer anonymous by default. They are a data subject. The language shifts quietly, but the implications are significant: what once appeared as incidental presence now carries traceability. The camera no longer captures a passerby. It captures a profile. From Eye to Scanner The street is no longer a neutral stage. Today, a camera in a public space functions less like an artist's eye and more like a biometric scanner. In many jurisdictions, the moment a face is captured, it ceases to be "character" and becomes "personal data." Once an image is uploaded, licensed, or even used to build a photographer's online brand, it enters a commercial context that strips away the old protections of "artistic freedom." This isn't just a theoretical threat. Reverse-image search has turned every street portfolio into a searchable database. A stranger in your frame is now one click
Apr 9, 2026 · via fstoppers.com
Abstract Image restoration is a vital research area in computer vision, focusing on reconstructing high-quality clear images from degraded observations. Common types of degradation include noise and blur, which may stem from imaging device limitations, environmental interference, and other factors. This paper centers on the design and optimization of multi-stage image restoration networks, conducting in-depth exploration of feature extraction, feature fusion, attention mechanisms, and their practical applications. A multi-stage hybrid attention mechanism-based image restoration network is proposed. Initially, each stage progressively extracts and restores image features. Then, an adaptive feature fusion block enables effective cross-stage information transfer. Finally, by calculating losses at each stage and assigning different weights, the network achieves stable convergence during training. The hybrid attention mechanism enhances the model’s focus on critical features and improves its understanding of the overall image structure. Outstanding performance has been achieved in both image deblurring and denoising tasks. On the GoPro dataset, the restored results achieved a PSNR of 33.26 and an SSIM of 0.963. On the SIDD dataset, the restored results reached a PSNR of 40.23 and an SSIM of 0.963. Furthermore, ablation experiments demonstrated the effectiveness of the multi-stage model, hybrid attention mechanism, and adaptive feature fusion block. Similar content being viewed by others Data availability The datasets generated and/or analysed during the current study are available in the public repositories listed below. For image de-blurring: the GoPro blur dataset (3214 images, 1280 \(\times\) 720 px) was downloaded from https://github.com/SeungjunNah/DeepDeblur_release. The originally provided train/test split (2103/1111 images) was adopted. High-resolution images were cropped into 512 \(\times\) 512 px patches to accelerate training and inference. For image de-noising: the Smartphone Image Denoising Dataset (SIDD) was obtained from https://www.eecs.yorku.ca/ kamel/sidd/. It contains 31888 noisy/clean image pairs; we used the standard split (30 608 training and 1 280 validation images) after per-image
Apr 9, 2026 · via nature.com
A field built long before the boom Artificial intelligence refers to methods that let machines perform tasks linked to human cognition, including pattern recognition, learning, and decision-making. Early systems relied on rule-based programs, which engineers designed to follow fixed instructions in controlled environments. While those systems worked well within narrow limits, they broke down once inputs became unpredictable. As researchers confronted those limits, they shifted toward machine learning in the late 20th century, training algorithms on large datasets rather than encoding rules directly. This transition improved performance in areas like speech recognition and image classification. As computing power expanded, models grew more complex, which in turn allowed deep learning systems to process raw data through layered neural networks. Cycles that shaped today’s moment Although progress appears steady in hindsight, AI research has moved through repeated cycles of optimism and retreat. Funding surged when results looked promising, then declined when systems failed to meet expectations. During those quieter periods, researchers refined techniques and built the groundwork for later advances. In that context, the current wave looks less like a sudden break and more like an acceleration of earlier work. Generative models reached the public after years of incremental gains in data processing and neural network design. Their rapid adoption reflects not only technical progress but also sustained corporate investment. Generative systems under pressure Generative AI systems produce text, images, and code by modeling statistical relationships within massive datasets. Rather than reasoning through problems, they predict outputs that match patterns seen during training. This approach allows them to generate fluent language and detailed content across many domains. At the same time, that strength exposes a consistent weakness. These systems can produce errors that sound convincing, which has raised concerns in fields where accuracy is paramount. Researchers continue to test reliability in areas
Apr 9, 2026 · via futura-sciences.com
Wladimir Silva and colleagues at North Carolina State University have developed an adaptive quantum algorithm for matrix multiplication that reduces the complexity of inner product calculations to O (log N) by using Quantum Random Access Memory. Their new “Adaptive Stacking” framework dynamically adjusts the algorithm’s execution, enabling compatibility with both near-term and fault-tolerant quantum systems and offering a flexible range of time complexities. Validation through a Quantum Machine Learning simulation on the MNIST dataset shows 96% accuracy and suggests a pathway towards sharply more efficient high-dimensional linear algebra operations. Demonstrated numerical stability and MNIST classification accuracy with a dynamically adjusted AQ-Stacker, a new hybrid quantum-classical algorithm, achieves 96 per cent accuracy on the MNIST dataset, exceeding the performance of previous methods with improved numerical stability in a practical quantum machine learning simulation. Realising super-classical efficiency in high-dimensional linear algebra has long been hindered by the difficulty of reconciling quantum speedups with the limitations of current quantum hardware. The MNIST dataset, comprising 70,000 labelled images of handwritten digits, serves as a standard benchmark for evaluating machine learning algorithms, particularly those dealing with image recognition and classification. Achieving high accuracy on this dataset demonstrates the algorithm’s ability to handle complex data representations and perform meaningful computations. AQ-Stacker overcomes this challenge by dynamically adjusting its execution pattern to optimise performance. This dynamic adjustment is crucial because the performance of quantum algorithms is heavily influenced by the number and quality of available qubits, as well as the coherence time, the duration for which qubits maintain their quantum state. Quantum Random Access Memory reduces the complexity of computing vector inner products to O(log N), enabling a tunable time-complexity range potentially reaching O(N2) on fault-tolerant systems. The use of QRAM allows for the preparation of quantum states in O(log N) time, significantly reducing data loading bottlenecks
Apr 9, 2026 · via quantumzeitgeist.com
This $99.99 AI tool combines writing, images, and more in one app TL;DR: 1min.AI is on sale for $99.99 (reg. $540), offering a lifetime subscription that combines multiple AI models and tools for writing, image generation, document processing, and more in one platform. Using AI efficiently often means using more than one tool, which is where things can start to feel a bit counterproductive. 1min.AI takes a more streamlined approach by combining those functions into a single platform. For a limited time, its lifetime plan is on sale for $99.99 (reg. $540). 1min.AI brings together multiple AI models like GPT, Claude, Gemini, MistralAI, and more into one workspace, so you can handle different tasks without switching between apps. Whether you’re writing, brainstorming, generating visuals, or analyzing documents, it’s designed to keep everything in one place. Instead of focusing on just one function, the platform covers a wide range of everyday use cases. For writing and content tasks, it can generate blog posts, rewrite drafts, expand or shorten content, and help with keyword research and SEO. It also includes tools for paraphrasing, summarizing, and even adapting content for platforms like LinkedIn or X, which can be useful if you’re managing multiple channels. On the visual side, it offers image generation and editing tools, including background removal, object editing, and upscaling. You can also generate images from prompts or sketches, adding a layer of flexibility to creative projects. For document-heavy work, it supports summarizing, translating, and interacting with files like PDFs and presentations. You can ask questions about documents or extract key information without manually digging through them. It also extends into audio and video, with features like text-to-speech, speech-to-text, voice tools, and basic video generation or editing, making it useful if your workflow includes multimedia content. Of course, like most AI
Apr 9, 2026 · via mashable.com

Constant Dullaart - The European Classes - Goethe-Institut United Kingdom
United Kingdom
Search
Constant Dullaart The European Classes
Leighton, Room 107
As artificial intelligence progresses, so-called âconvolutional networksâ (ConvNets) now exist that can recognise objects within photographic images. Constant Dullaart has retrained these image recognition networks to focus purely on European images, creating an image dataset formed in collaboration with universities across Europe. He asks, âHow can Europeâs diverse cultural entities be represented within these networksâ capacity to recognise, create and interact?â The outcome will consist of terabytes of images that Dullaart hopes will represent the European collective identity in 2017.
Constant Dullaart was born in the Netherlands. He reflects on the cultural and social effects of communication and image processing technologies. His work includes websites, performances, fake armies and manipulated found images, presented both offline and in the public space of the Internet.
Apr 9, 2026 · via goethe.de
Notes from the Asia-Pacific region: Robust conversations at richly rewarding Summit The IAPP Global Summit 2026 brought together diverse voices — from scientists and legal scholars to authors and global figures — to discuss privacy as a deeply human, culturally significant and increasingly urgent challenge. Contributors: Charmian Aw AIGP, CIPP/A, CIPP/E, CIPP/US, CIPM, FIP Partner Hogan Lovells Editor's note The IAPP is policy neutral. We publish contributed opinion pieces to enable our members to hear a broad spectrum of views in our domains. This year's IAPP Global Summit 2026, held just last week, featured a series of robust, potent conversations, which lingered long after attendees departed. Highlights continuing to brew in my mind are distilled below. The conference opened with cognitive scientist Maya Shankar, whose work and book on decision-making invite us to reconsider the invisible forces shaping human behavior. Shankar's keynote was nothing short of a hand-crafted, house-blend reflection on how small interventions can stir profound change. In a conversation with IAPP Vice President and Chief Knowledge Officer Caitlin Fennessy, CIPP/US, world-renowned author Salman Rushdie added a deeply human and cultural dimension. Reflecting on his experiences and observations, including life in India, Rushdie highlighted how privacy is neither universal nor evenly distributed. In the slums, privacy is a luxury. Cultural norms, social structures and economic realities shape what privacy means — and whether it is even attainable. Boston University School of Law Professor Woodrow Hartzog's keynote was a calibrated dark roast — deep, complex and intentionally uncomfortable. He spoke candidly about privacy's ongoing identity crisis. Against the backdrop of shrinking budgets and expanding technological capability, Hartzog warned of the risks posed by surveillance systems, including facial recognition, and the creeping normalization of intrusive technologies. Unfettered artificial intelligence, he argued, threatens not just data protection, but the very fabric of
Apr 9, 2026 · via iapp.org
Abstract Automatic and high-precision video-based animal re-identification (Re-ID) has become an important tool for wildlife conservation and behavioral research. However, state-of-the-art image-based methods, while effective on static images, fail to make full use of the rich dynamic information in videos, such as gait and motion patterns. To bridge this gap, we propose HST-Former, a novel framework that extends data efficiency and species-agnostic principles into the temporal domain. The core innovation of this method is the Hierarchical Spatio-Temporal Transformer Aggregator (HSTTA)—a customized transformer architecture designed to process and integrate all local features from an entire animal trajectory. By modeling both spatial and temporal feature dependencies, HSTTA learns complex long-range relationships and produces a single, highly discriminative video-level descriptor. The hierarchical design first summarizes intra-frame features and then aggregates them across frames, effectively addressing the computational challenges of standard transformers on long sequences. In addition, we introduce a spatio-temporal consistency constraint to enhance the geometric verification step, improving re-ranking accuracy. Our model significantly outperforms current state-of-the-art baselines. To validate its effectiveness, we conduct comprehensive evaluations on three public datasets, where HST-Former achieves the best performance across all key metrics, including Top-1, Top-3, and Top-5. Similar content being viewed by others Data availability The datasets used during the current study are available from the corresponding author on reasonable request. References Araujo, A. et al. Recurrent neural networks for person re-identification revisited. IEEE Trans. Multimed. 21, 2813–2824 (2019). Whytock, R. C. et al. Optimizing the automated recognition of individual animals to support population monitoring. Ecol. Evol. 13, e10210 (2023). Crouse, D. et al. Identification of animal individuals using deep learning: A case study of the giant panda using infrared camera images. Biol. Conserv. 242, 108405 (2020). Francis, D. Review of the current research in animal individual recognition. Modern Sci. 1, 80–85 (2023). Clapham, M.,
Apr 9, 2026 · via nature.com
Abstract This work introduces SqueezeViT (Squeeze Vision Transformers), a compact yet effective architecture based on Vision Transformers (ViT) designed for chest X-ray (CXR) image classification. In contrast to traditional ViT architectures, which are computationally demanding, SqueezeViT employs a novel squeezing procedure that effectively lowers token dimensions without compromising important visual components, leading to expedited inference and decreased memory consumption. The designed model is tested for two commonly used public datasets, NIH Chest X-ray and CheXpert, providing a diverse range of thoracic pathologies. SqueezeViT reduces the number of parameters by 43.2% compared to the baseline MobileViT1, and up to 95.4% compared to other state-of-the-art (SOTA) models. The suggested model offers up to 16.5% improvement in the area under the receiver operating characteristic curve (AUROC) compared to SOTA models, and it is, in general, superior to the baseline and effective convolutional neural networks CNNs2 in numerous tasks. Such developments make the proposed SqueezeViT approach an attractive option for a wide variety of applications. The findings indicate that SqueezeViT outperforms the current SOTA classifiers while maintaining a lightweight model architecture. In turn, such results emphasize the possibilities of using SqueezeViT in real clinical environment, where the amount of computational resources can be constrained. Similar content being viewed by others Data availability The datasets used in this work are available at: [https://www.kaggle.com/datasets/nih-chest-xrays/data](https:/www.kaggle.com/datasets/nih-chest-xrays/data)[https://aimi.stanford.edu/datasets/chexpert-chest-x-rays](https:/aimi.stanford.edu/datasets/chexpert-chest-x-rays)The model of the proposed SqueezeViT is available on the link: [https://github.com/vijay13787/SqueezeVIT.git](https:/github.com/vijay13787/SqueezeVIT.git). References Wang, X. et al. ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases. in. IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2017. (2017). Haritha, D., Pranathi, M. K. & Reethika, M. COVID detection from chest X-rays with DeepLearning: CheXNet. in 5th international conference on computing, communication and security (ICCCS). 2020. IEEE. 2020. IEEE. (2020). Hasanah, U. et al. CheXNet and feature
Apr 9, 2026 · via nature.com
European Union’s Digital Omnibus Package Heats Up the Debate on Human Rights Protection - Human Rights Research Center - 8 hours ago - 6 min read Author: Ana Budeanu April 9, 2026 HRRC acknowledges the efforts of the European Union to establish a more cohesive set of digital rules across its territory, but emphasises that any policy must adhere to the principles found in the international human rights body of laws. The European Union (EU) Digital Omnibus Package has member states and human rights organizations in the EU debating whether this set of legislative acts is a good choice for EU citizens, especially those in vulnerable groups such as people with disabilities or members of LGBTQIA+, as they seem more likely to diminish already established rules that protect privacy, prevent discrimination, and ensure accountability in digital systems. To date, the European Commission has introduced the “Digital Package,” a comprehensive regulatory framework encompassing both the Digital Omnibus Regulation and the parallel Digital Omnibus on Artificial Intelligence Regulation. The EU’s digital regulatory construction is currently undergoing a significant change through the Digital Omnibus Package, which has two main acts: the Digital Omnibus on AI and the Digital Omnibus Regulation. The package is intended to help the EU’s digital economy and lessen administrative pressures on businesses by creating a more cohesive set of rules for everything which takes place online, from purely business transactions to communications, aiming to resolve legal uncertainties within the complex web of the General Data Protection Regulation (GDPR), the Data Act, and the ePrivacy Directive. The more important of the two acts is the EU Digital Omnibus on AI proposal, a legislative initiative seeking to simplify and modernize the existing regulations for data protection, artificial intelligence, and cybersecurity within broader online activities, including business, education or recruitment. The starting
Apr 9, 2026 · via humanrightsresearch.org
Abstract Image defect detection of power transmission and transformation equipment is a key technology in the operation and maintenance of power systems, which is of vital importance for ensuring the safe and stable operation of the power grid and enhancing its self-healing capacity. However, this task has long been confronted with the core challenge of a limited number of defect samples for specific categories. Traditional deep learning models trained on large-scale data find it difficult to achieve accurate detection under small-sample conditions. To address this issue, this study proposes a lightweight segmentation network based on the Mamba architecture, Residual Mamba (ResMamba). This network adopts a six-level U-shaped codec structure and innovatively integrates the Mamba module into the visual state space (VSS) as the codec link. Its core ResVSS module significantly reduces the number of parameters by removing a redundant linear layer (accounting for 38% of the original VSS module parameters) in the internal shortcut connection of the original VSS module, and introduces deep convolutional blocks and learnable scale parameters to dynamically scale residual connections. It enhances feature representation ability while reducing model complexity. In addition, the skip connection part introduces a multi-level and multi-scale information fusion mechanism, effectively integrating cross-scale features by generating spatial and channel attention maps, thereby enhancing the model’s performance in multi-specification defect detection. Experimental results on public datasets (including specialized small-sample settings) show that ResMamba achieves superior segmentation accuracy while maintaining a low parameter count, effectively balancing computational efficiency and detection performance. It outperforms both general segmentation models and domain-specific models for power equipment defect detection, providing a reliable new solution for small-sample defect detection of power equipment and enhancing the self-healing ability of power grids. Similar content being viewed by others Data availability The data cannot be shared publicly because it contains sensitive information/is subject
Apr 9, 2026 · via nature.com
Woodchuck Named “Waste Diversion Solution of the Year” In 2026 CleanTech Breakthrough Awards Program Published Thursday, April 9, 2026 | 5:01 a.m. Updated Thursday, April 9, 2026 | 5:03 a.m. GRAND RAPIDS, Mich.--(BUSINESS WIRE)--Apr 9, 2026-- Woodchuck , the AI-powered climate-tech startup redefining how construction and manufacturing industries handle wood waste, today announced it has been awarded “Waste Diversion Solution of the Year” in the 3 rd annual CleanTech Breakthrough Awards program conducted by CleanTech Breakthrough , a leading independent market intelligence organization that evaluates and recognizes standout climate and clean technology companies, products and services around the globe. This press release features multimedia. View the full release here: https://www.businesswire.com/news/home/20260409517100/en/ Woodchuck was recognized for its AI-enabled waste diversion platform that transforms wood waste into renewable energy by combining AI-driven technology, on-site logistics, and bioenergy partnerships to deliver a scalable, measurable solution that turns waste into clean baseload power. Woodchuck deploys AI-enabled image recognition directly on job sites. Instead of relying on manual sorting, Woodchuck integrates smart containers, sensor technology, and computer vision systems that identify, classify, and track wood waste in real time. This increases diversion rates, reduces contamination, reduces landfill dependency while supplying regional bioenergy facilities with renewable feedstock and generates transparent, auditable data for contractors and project owners. Woodchuck’s model helps contractors routinely achieve 30–40% cost savings compared to traditional waste hauling while meeting aggressive ESG and diversion targets. Waste becomes a measurable sustainability advantage to help companies win bids on high-profile projects, including EV battery plants and hyperscale AI data centers. “We’re not just simply diverting waste, but building a circular system where technology, economics, and climate impact reinforce one another. Our solution turns sustainability into cost savings and demonstrates that climate leadership can be both profitable and scalable. Woodchuck’s impact also extends beyond carbon accounting into
Apr 9, 2026 · via lasvegassun.com
For those of you just tuning in, I’ll repeat that a little over a year ago I made a visit to the archives belonging to the Missouri History Museum in Forest Park. There I met the archivist Lauren Sallwasser who it turned out is a fellow Maplewoodian. She showed me their collection of documents and photographs that had been donated by Margaret Sappington Townsend. She was kind enough to provide me with digital copies of two of the photo albums from that collection. It was a mind-blowing experience. I’ve been trying to get back ever since. Although, I have posted many of the images from those two albums that I consider solid gold nuggets from the rich vein of Maplewood history. There are many more that are worth a look. Last post we took a look at the first ten pages from Album One Click on this if you would like to see the First Ten Pages from Album One again. Still from Album One, here are ten more pages beginning with page 11. Page eleven. The address on this house is very clearly 7177. From other images in this collection, I have learned that it was on Manchester, west of Sutton even though the address would have you think otherwise. Images added to the album later show that the address had been changed to 7477. This is a fairly common problem for historians. As the population increased, it was sometimes necessary for the post office to change the addresses. Page twelve. Philip’s page. Since I have had much enjoyment from having the whole of the two albums to examine, I thought that the only way for the followers of this space to experience that was to post all of the images contained within the albums. That is a tall
Apr 9, 2026 · via 40southnews.com
Abstract Ultrasound imaging is a crucial modality for breast cancer diagnosis, with lesion segmentation being an essential component of computer-aided diagnosis systems. However, the high cost and effort associated with pixel-level annotation significantly hinder the advancement of breast ultrasound (BUS) image segmentation algorithms. In this paper, we propose a multi-level semi-supervised generative adversarial network (MLSS-GAN) that incorporates a generator and a novel multi-level classifier acting as the discriminator. While some previous approaches have incorporated multi-level adversarial learning, our discriminator employs an encoder for feature extraction and two decoders designed for distinct image-level and pixel-level adversarial tasks, enabling more comprehensive segmentation supervision. Specifically, one decoder performs image-level classification, while the other conducts pixel-level classification. This multi-level adversarial strategy compels the generator to produce images that closely resemble real BUS images at different feature representations, thereby enriching the classification feature space. To assess the effectiveness of the proposed method, comprehensive comparisons with both fully supervised and semi-supervised approaches are conducted on the BUSI dataset. Experimental results demonstrated the superior performance of our method, especially under conditions with only a limited number of annotated BUS images. Similar content being viewed by others Data availability The BUSI dataset analyzed during the current study is available in the BUSI repository, https://doi.org/10.1016/j.dib.2019.104863. References Bray, F. et al. Global cancer statistics 2022: Globocan estimates of incidence and mortality worldwide for 36 cancers in 185 countries. CA Cancer J. Clin. (2021). Masood, S., Sharif, M., Masood, A., Yasmin, M. & Raza, M. A survey on medical image segmentation. Curr. Med. Imaging 11, 3–14 (2015). Gu, Y. et al. Deep learning based on ultrasound images assists breast lesion diagnosis in china: a multicenter diagnostic study. Insights Imaging 13, 124 (2022). Ranjbarzadeh, R. et al. Breast tumor localization and segmentation using machine learning techniques: Overview of datasets, findings, and methods.
Apr 9, 2026 · via nature.com
U.S. Air Force Senior Master Sgt. Matthew Broussard, front right, Air Force Global Strike Command fire and emergency services program manager, receives a medallion from his wife as part of the Chief Recognition Ceremony at Barksdale Air Force Base, Louisiana, April 3, 2026. The ceremony is held to recognize Airmen who have achieved the rank of chief master sergeant. (U.S. Air Force photo by Senior Airman Rhea Beil)
| Date Taken: | 04.03.2026 |
| Date Posted: | 04.08.2026 17:28 |
| Photo ID: | 9603888 |
| VIRIN: | 260403-F-DY500-1009 |
| Resolution: | 8202x5468 |
| Size: | 4.61 MB |
| Location: | BARKSDALE AIR FORCE BASE, LOUISIANA, US |
| Web Views: | 12 |
| Downloads: | 0 |
This work, Chief Recognition Ceremony 2026 [Image 8 of 8], by SrA Rhea Beil, identified by DVIDS, must comply with the restrictions shown on https://www.dvidshub.net/about/copyright.
Apr 9, 2026 · via dvidshub.net
Scientists develop three-in-one diode PN junction diodes are the underpinning of modern optoelectronic systems. But, typically, they perform only a single task. To process complex visual signals, optical imaging systems often rely on the introduction of a third electrode or integrating diodes with external logic circuits. But these architectures increase hardware complexity and bring von Neumann issues — the physical separation of sensing, memory, and computing—leading to larger areas and increased power consumption. There is a lack of a compact, single-device solution. Now, in a study titled 'A single diode with integrated photosensing, memory and processing for neuromorphic image sensors' published in Nature Electronics, a research team led by SUN Haiding from the University of Science and Technology of China of the Chinese Academy of Sciences, along with the collaborators from domestic and foreign universities, have developed a multifunctional photodiode architecture. They appraoch, they say, provides a hardware solution for low-power neuromorphic visual cameras with integrated sensing, memory, and computing, breaking through the traditional semiconductor physical limit of unidirectional conductivity. The new PN junction design is based on energy band engineering. The team constructed high-crystal-quality vertical GaN-based PN junction diode arrays on conductive silicon substrates. By inserting a wide-bandgap n-AlGaN layer into the GaN-based PN junction, they formed a localised 'charge storage layer' internally via energy band bending. This embedded structure endowed the device with unique charge trapping and releasing capabilities, enabling arbitrarily mode switch through simple external bias voltages without altering the device structure. The team demonstrated that a single device could exhibit three independent and freely switchable working modes: photosensing, photosynapse, and photomemory. Under zero bias, the device exhibited stable self-driven photosensing characteristics for real-time image sensing. Under a constant bias, it demonstrated photosynaptic behaviour modulated by pulse frequencies. Under pulsed bias operation, it achieved multi-state optical memory
Apr 9, 2026 · via compoundsemiconductor.net