α,α-Dideuterio benzyl alcohols are important building blocks for the synthesis of deuterium-labeled medicines and agrochemicals. We have developed the first general single-electron transfer reductive deuteration of readily commercially available aromatic esters for the synthesis of α,α-dideuterio benzyl alcohols using benign D₂O and a mild single-electron donor SmI₂. This operationally convenient method features very good functional group tolerance and high deuterium incorporations (>95% D₂). The potential impact has been exemplified by the synthesis of numerous deuterium labeled building blocks of important bioactive compounds. Most crucially, the method represents the first example of selective reductive deuteration of benzylic-type ketyl radicals using mild and highly chemoselective lanthanide(II) reagents. Received: 28 August 2020 Accepted after revision: 17 September 2020 Article published online: 16 October 2020 © 2020. Thieme. All rights reserved Georg Thieme Verlag KG Rüdigerstraße 14, 70469 Stuttgart, Germany

Maize aerial root architecture is a key trait for assessing yield stability due to their critical role in biological nitrogen fixation and mechanical lodging resistance. However, automated identification in field environments remains challenging due to small targets, severe occlusion, and the limitations of traditional destructive measurement methods. To address these issues, we initially established a high-definition maize video dataset (380 GB, 1900 mins) using the Maizrob, and subsequently constructed the first standardized aerial root image library (12 million images) to fill the domain gap. Based on these datasets, a hierarchical identification framework was developed, utilizing YOLOv10 for feature extraction and a novel Occlusion-Aware Multi-Frame Fusion (OMF) strategy for decision-making. This strategy synthesizes multi-view information to specifically mitigate the impact of cluttered backgrounds and visual blocking. Experimental results demonstrate that the OMF strategy significantly improves robustness, raising Cohen's Kappa coefficient from 0.34 to 0.64 in ablation studies. Consequently, the overall framework achieves a plant-level accuracy of 90.7% in complex field scenarios. This performance represents a substantial improvement over traditional single-frame detection models (65.6%-75.5%), confirming the effectiveness of the dataset-supported cascade architecture. This study provides an accurate, non-destructive technical solution for maize phenotyping and offers a reference for the automatic identification of local micro-traits in other crops.

Generalized few-shot semantic segmentation (GFSS), which requires strong segmentation performance on novel classes while retaining the performance on base classes, is attracting increasing attention. Recent studies have demonstrated the effectiveness of applying visual prompts to solve GFSS problems, but there are still unresolved issues. Due to the confusion between the backgrounds and novel classes foreground during base class pre-training, the learned base visual prompts will mislead the novel visual prompts during novel class fine-tuning, leading to sub-optimal results. This paper proposes a background-enhanced visual prompting Transformer (Beh-VPT) to solve the problem. Specifically, we innovatively propose background visual prompts, which can learn potential novel class information in the background during base class pre-training and transfer the information to novel visual prompts during novel class fine-tuning via our proposed Hybrid Causal Attention Module. Additionally, we propose a background-enhanced segmentation head that is used in conjunction with background prompts to enhance the model’s capacity for learning novel classes. Considering the GFSS settings that take into account both base and novel classes, we introduce Singular Value Fine-Tuning in the non-meta learning paradigm to further unleash the full potential of the model. Extensive experiments show that the proposed method achieves state-of-the-art performance for GFSS on PASCAL-$5^{\mathrm{i}}$ and COCO-${20}^{\mathrm{i}}$ datasets. For example, considering both base and novel classes, the improvements in mIoU range from 0.47% to 1.08% (COCO-${20}^{\mathrm{i}}$) in the one-shot and five-shot scenarios, respectively. In addition, our method does not cause a fallback of mIoU in base classes relative to the baseline.