Matthew Davis

Belgium came into the World Cup ranked No. 10 in the world, and the team has everything on the line after Sunday's scoreless tie with Iran.

A former Minnesota Vikings star is painfully familiar with what North Dakota State did to his alma mater, Iowa.

Picture this: a self-driving truck rolls into a remote mine site at dusk, its headlights cutting through dust swirling off the haul road. The vehicle's AI needs one thing above all else—an accurate, continuous map of every access track, switchback, and turnaround, not just a smudge of pixels where the road *might* be. Without that topological fidelity, the truck might stall at a cul-de-sac or take a detour that never existed, costing hours and thousands of dollars in lost productivity. Now imagine if the same AI could ingest a single high-resolution satellite image and, in seconds, deliver a vectorized road network that preserves every junction and branch—no gaps, no phantom shortcuts. That future is closer than you think, thanks to work led by Zhengbo Yu at the School of Earth Science and Resources, China University of Geosciences (Beijing).

Algeria bounced back and Jordan saw its World Cup debut turn sour on Monday in Santa Clara.

As North Dakota State gets ready for its first FBS season, the Bison have a significant announcement for fans.

Zhen Cao, a researcher at Guilin University of Technology in China, has developed a groundbreaking tool for monitoring floating raft aquaculture (FRA) in coastal waters—a challenge that has long bedeviled marine planners and energy companies operating in these zones. His innovation, FALOD (Frequency domain enhanced Lightweight Oriented Detector), uses advanced image processing to detect and track aquaculture rafts in high-resolution satellite imagery with remarkable precision. What makes FALOD special isn't just its accuracy—it's also fast and efficient, making it suitable for large-scale, long-term monitoring.

A team of researchers has developed a framework that could transform how long-form action quality is assessed in high-stakes environments, where precision and reliability are non-negotiable. The system, called Multi-Level Cue Refinement (MLCR), breaks away from traditional one-size-fits-all modeling of video sequences by organizing quality evidence in three distinct layers. This structured approach enables the extraction of meaningful signals from minutes-long audiovisual recordings without losing critical cues amid noise, redundancy, or temporal distortion. At its core, MLCR treats long-term action quality assessment not as a simple scoring problem, but as a carefully orchestrated process of isolating, combining, and refining evidence across modalities such as vision, motion, and audio.

In the rugged landscapes of South Africa's Limpopo Province, where the earth's crust tells a story of ancient geological upheavals, a quiet revolution is unfolding—one that could redefine how we hunt for magnesite, a mineral as critical to industry as it is elusive. Led by L.D. Ndou, a researcher at the University of Venda, a new study has harnessed the power of satellite imagery to uncover hidden deposits of this valuable mineral, offering a cost-effective and rapid method for exploration that could reshape the energy and refractory sectors.

Researchers at CUG have just unveiled a neural architecture that could have implications far beyond the radiology department. Their Uncertainty-Aware Hypergraph Refinement Network, UHR-Net, is built to solve the two most persistent headaches in medical image segmentation: the way faint lesions bleed into healthy tissue and the way tiny, early-stage lesions disappear inside multi-scale feature maps. The team's solution couples a geometry-aware copy-paste augmentation with hard-negative mining to teach the model to distinguish ambiguous boundaries before fine-tuning even begins. In effect, UHR-Net learns to treat uncertain background regions as adversarial samples, forcing its instance encoder to sharpen its discrimination for small or poorly defined lesions.

In a groundbreaking study that could redefine how coastal regions are monitored and managed, researchers have demonstrated the superior accuracy of AlphaEarth Foundations (AEF) data in tracking coastline and land-use changes in China's Pearl River Estuary (PRE). The findings, published in *Remote Sensing* (远程传感), highlight how this advanced data source outperforms traditional Sentinel-2 imagery, offering unprecedented precision in environmental monitoring—a critical tool for industries like energy, where coastal stability and land-use planning directly impact operations and sustainability.