Even a toddler knows that plants need water. It’s perhaps the first thing we learn about these green lifeforms. But how plants budget this resource varies considerably. The kapok trees of the Amazon have adopted vastly different strategies than the switchgrass of the American plains. Unfortunately, it’s hard to directly measure which ones prevail in different ecosystem types and how they shift under changing conditions.

Industrial anomaly detection is crucial for maintaining quality control and reducing production errors, but traditional supervised models require extensive datasets. While embedding-based methods are promising for unsupervised anomaly detection, they are highly memory-intensive and unsuited to low-light conditions. In a new study, researchers developed a new unsupervised model that utilizes both well-lit and low-light images to achieve computationally efficient and memory-friendly industrial anomaly detection.

A research team led by Professor Jie Zeng and Associate Researcher Han Yan from the University of Science and Technology of China, in collaboration with Professor Chao Ma from Hunan University, has developed a novel ceria-supported platinum bilayer cluster catalyst. This breakthrough material demonstrates exceptional catalytic activity and stability in alkene hydrosilylation reactions while achieving atomic-level precision structural identification of the catalyst.