[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-3256":3,"related-3256":59},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":6,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":24,"tags":26,"search_phrases":32,"slug":35,"view_count":36,"doi":37,"paper":38,"created_at":58},3256,"Leakage-aware, calibrated, and explainable deep learning for robust almond disease classification","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112464","Leakage-aware, calibrated, and explainable deep learning for robust almond disease classification。Computers and Electronics in Agriculture","面向稳健杏仁病害分类的泄漏感知、校准且可解释的深度学习。《农业计算机与电子》",null,"Computers and Electronics in Agriculture","2026-09-22T00:00:00Z","论文",10,false,72,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},12,20,17,14,9,1,"核心期刊论文，方法上有防泄漏、校准与可解释性创新，但作物小众、属细分技术进展，未达每日精选门槛。",[25],{"name":10,"url":6},[27,28,29,30,31],"智慧农业","农业人工智能","深度学习","病害识别","巴旦木",[33,34],"巴旦木 病害 深度学习","农业人工智能 智慧农业 深度学习 病害识别","巴旦木病害深度学习-3256",0,"10.1016\u002Fj.compag.2026.112464",{"doi":37,"openalex_id":39,"authors":40,"venue":10,"cited_by_count":36,"oa_url":9,"card":51,"direction":55,"ingested_from":57},"W7213988471",[41,44,47,49],{"name":42,"orcid":43},"Abebaw Degu Workneh","https:\u002F\u002Forcid.org\u002F0000-0001-7694-1577",{"name":45,"orcid":46},"Badr Elkari","https:\u002F\u002Forcid.org\u002F0000-0002-0893-783X",{"name":48,"orcid":9},"Meryam El Mouhtadi",{"name":50,"orcid":9},"Mohammad Furqan Ali",{"tldr":52,"method":53,"finding":54,"direction":55,"opportunity":56},"提出防泄漏、校准且可解释的深度学习框架，用于稳健的杏仁病害分类。","采用防数据泄漏的深度学习训练、概率校准与可解释性分析。","该框架能提升杏仁病害分类的稳健性、可信度与可解释性。","农业人工智能与决策模型","可探索防泄漏与校准机制在其他作物病害识别中的泛化及田间部署。","openalex","2026-09-23T23:30:01.628054Z",{"total":60,"page":22,"page_size":60,"items":61},6,[62,104,151,193,235,291],{"id":63,"title":64,"url":65,"summary":66,"summary_zh":67,"content":9,"source_name":68,"source_url":65,"published_at":69,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":70,"score_detail":71,"sources":76,"tags":78,"search_phrases":80,"slug":83,"view_count":36,"doi":84,"paper":85,"created_at":103},2910,"Precision Diagnosis of Apple Leaf Diseases Across Infection Stages via Web-Based Analysis","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs11119-026-10451-5","Abstract Purpose Throughout the growing season, leaf diseases pose a huge risk to apple fruit yield and quality. While early detection targets timely intervention, understanding disease progression across different stages is essential for effective treatment and resistance breeding. Methods This study proposes a deep learning-based detection system that identifies apple leaf diseases at various infection stages in real-world environments. Most existing methods have focused on disease detection in controlled conditions, while the developed system operates in complex field settings with variable lighting and background noise. Two YOLOv8-based variants integrated with a multi-head self-attention (MHSA) module were developed to improve detection of dense, tiny, and feature-similar lesions. In addition, a web-based interactive tool was built by combining the Segment Anything Model (SAM) with the detection model to support leaf isolation, disease localization, infection estimation, severity categorization, and automatic report generation. Results The training results on proposed PA-ALeaf dataset demonstrated that YOLOv8-MHSA_b_h varaint achieved the highest performance (Precision 60.5%, mAP50 56.3%, F1 confidence 56.0%), while maintaining real-time inference (12.38 ms per image), compared to state-of-the-art models. Conclusion By bridging practical disease monitoring and scientific research, our system offers a comprehensive, scalable solution for apple leaf disease detection in real-world orchards. This system could benefit both farmers by enabling multi-stage intervention and researchers by providing insights into disease progression.","摘要 目的 在整个生长季中，叶片病害对苹果果实产量和品质构成巨大风险。虽然早期检测有助于及时干预，但了解病害在不同阶段的进展对于有效治疗和抗性育种至关重要。方法 本研究提出了一种基于深度学习的检测系统，可在真实环境中识别不同感染阶段的苹果叶片病害。现有大多数方法侧重于受控条件下的病害检测，而所开发的系统可在光照变化和背景噪声复杂的田间环境中运行。研究开发了两种基于YOLOv8的变体，并集成了多头自注意力（MHSA）模块，以提高对密集、微小和特征相似病斑的检测能力。此外，通过将分割一切模型（SAM）与检测模型相结合，构建了一个基于网络的交互式工具，以支持叶片分离、病害定位、感染估计、严重程度分类和自动报告生成。结果 在所提出的PA-ALeaf数据集上的训练结果表明，与最先进的模型相比，YOLOv8-MHSA_b_h变体取得了最高性能（精确率60.5%，mAP50 56.3%，F1置信度56.0%），同时保持实时推理（每张图像12.38 ms）。结论 通过连接实际病害监测与科学研究，我们的系统为真实果园中的苹果叶片病害检测提供了一种全面、可扩展的解决方案。该系统既可使农民受益，实现多阶段干预，也可为研究人员提供病害进展的见解。","Precision Agriculture","2026-09-18T00:00:00Z",80,{"impact":72,"substance":73,"depth":72,"authority":20,"freshness":74,"relevant":22,"comment":75},18,22,8,"提出面向真实果园的多阶段苹果叶病检测系统并配套网页工具，方法新颖、数据与性能指标明确，对智慧植保具有参考价值。",[77],{"name":68,"url":65},[27,28,29,30,79],"苹果病害",[81,82],"苹果叶部病害 深度学习 识别","YOLOv8 苹果病害 检测","苹果叶部病害深度学习识别-2910","10.1007\u002Fs11119-026-10451-5",{"doi":84,"openalex_id":86,"authors":87,"venue":68,"cited_by_count":36,"oa_url":65,"card":98,"direction":55,"ingested_from":57},"W7213537759",[88,90,93,95],{"name":89,"orcid":9},"Kangrui Han",{"name":91,"orcid":92},"Hao Cai","https:\u002F\u002Forcid.org\u002F0000-0002-9879-2960",{"name":94,"orcid":9},"Kari Peter",{"name":96,"orcid":97},"Long He","https:\u002F\u002Forcid.org\u002F0000-0001-9781-6062",{"tldr":99,"method":100,"finding":101,"direction":55,"opportunity":102},"提出基于YOLOv8-MHSA与SAM的网页系统，实现苹果叶片病害多感染阶段实时检测与分级。","YOLOv8+多头自注意力，结合SAM分割，构建PA-ALeaf数据集与网页工具","YOLOv8-MHSA_b_h变体精度60.5%、mAP50 56.3%，单图推理12.38ms，可","可探索轻量化模型在移动端的部署，并融合时序数据预测病害发展轨迹。","2026-09-19T23:30:03.474350Z",{"id":105,"title":106,"url":107,"summary":108,"summary_zh":109,"content":9,"source_name":110,"source_url":107,"published_at":111,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":112,"score_detail":113,"sources":117,"tags":119,"search_phrases":121,"slug":123,"view_count":36,"doi":124,"paper":125,"created_at":150},1623,"Deep convolutional neural network-based automated identification and classification of mungbean foliar diseases","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffrai.2026.1848787","Early and accurate detection of plant diseases is critical in precision agriculture to improve crop management and yield. Mungbean ( Vigna radiata L.) is highly susceptible to several foliar diseases, including yellow mosaic, powdery mildew, leaf crinkle, and cercospora leaf spot, which cause substantial productivity losses. Despite expanding applications of deep learning in plant disease diagnosis, systematic multi-architecture evaluation for mungbean disease classification under natural field conditions remains limited. This study addresses this gap by evaluating five state-of-the-art deep convolutional neural network (DCNN) architectures on a large-scale, field-acquired mungbean dataset that captures real-world variability across environmental conditions and disease severity levels, distinguishing it from controlled laboratory studies. A total of 5,617 original images across five classes were used. Data augmentation was applied exclusively to the training subset after stratified splitting to prevent data leakage. The dataset was partitioned into training (70%), validation (15%), and testing (15%) subsets. VGG16, VGG19, ResNet50V2, DenseNet121, and InceptionV3 were evaluated using identical transfer learning and fine-tuning protocols. Model performance was assessed using AUC-ROC, Cohen's kappa coefficient, McNemar's test for pairwise statistical comparisons, five-fold cross-validation, and Grad-CAM-based interpretability. On the independent test set, InceptionV3 achieved the highest accuracy (98.47%) and macro-F1 (98.49%), followed by VGG16 (98.36%) and VGG19 (97.89%). AUC-ROC values exceeded 0.997 for all models, confirming excellent class discrimination. Grad-CAM visualizations further confirmed that model predictions were based on biologically relevant disease symptoms. The findings demonstrate the effectiveness of deep learning for robust disease recognition under realistic field conditions and highlight the potential of AI-based diagnostic tools for crop health monitoring, precision agriculture, and decision-support systems in mungbean production.","植物病害的早期准确检测对于精准农业中改善作物管理和提高产量至关重要。绿豆（*Vigna radiata* L.）极易感染多种叶部病害，包括黄花叶病、白粉病、叶皱缩病和尾孢叶斑病，这些病害会导致显著的生产力损失。尽管深度学习在植物病害诊断中的应用日益广泛，但在自然田间条件下对绿豆病害分类进行系统性多架构评估的研究仍然有限。本研究通过在大规模、田间采集的绿豆数据集上评估五种最先进的深度卷积神经网络（DCNN）架构来填补这一空白，该数据集捕捉了环境条件和病害严重程度方面的真实世界变异性，使其区别于受控实验室研究。共使用了涵盖五个类别的5,617张原始图像。在分层划分后，数据增强仅应用于训练子集，以防止数据泄漏。数据集被划分为训练集（70%）、验证集（15%）和测试集（15%）。VGG16、VGG19、ResNet50V2、DenseNet121和InceptionV3采用相同的迁移学习和微调协议进行评估。模型性能通过AUC-ROC、Cohen's kappa系数、用于成对统计比较的McNemar检验、五折交叉验证以及基于Grad-CAM的可解释性进行评估。在独立测试集上，InceptionV3取得了最高的准确率（98.47%）和宏平均F1分数（98.49%），其次是VGG16（98.36%）和VGG19（97.89%）。所有模型的AUC-ROC值均超过0.997，证实了优异的类别区分能力。Grad-CAM可视化进一步证实了模型预测基于生物学相关的病害症状。研究结果表明，深度学习在现实田间条件下实现稳健病害识别方面具有有效性，并凸显了基于人工智能的诊断工具在绿豆生产中的作物健康监测、精准农业和决策支持系统方面的应用潜力。","Frontiers in Artificial Intelligence","2026-09-03T00:00:00Z",74,{"impact":114,"substance":73,"depth":72,"authority":17,"freshness":115,"relevant":22,"comment":116},15,7,"系统评估多种深度学习模型在绿豆叶部病害识别中的性能，数据规模大且基于田间条件，具有较强应用价值。",[118],{"name":110,"url":107},[27,28,120,29,30],"绿豆",[34,122],"农业人工智能 智慧农业","农业人工智能智慧农业深度学习病害识别-1623","10.3389\u002Ffrai.2026.1848787",{"doi":124,"openalex_id":126,"authors":127,"venue":110,"cited_by_count":36,"oa_url":143,"card":144,"direction":149,"ingested_from":57},"W7207845216",[128,130,132,135,137,139,141],{"name":129,"orcid":9},"Shail Bala",{"name":131,"orcid":9},"SI Harlapur",{"name":133,"orcid":134},"Aditya Kamalakar Kanade","https:\u002F\u002Forcid.org\u002F0009-0004-8685-3326",{"name":136,"orcid":9},"M. P. Potdar",{"name":138,"orcid":9},"Gurupada Balol",{"name":140,"orcid":9},"VB Kuligod",{"name":142,"orcid":9},"Lingareddy Usha Rani","https:\u002F\u002Fwww.frontiersin.org\u002Fjournals\u002Fartificial-intelligence\u002Farticles\u002F10.3389\u002Ffrai.2026.1848787\u002Fpdf",{"tldr":145,"method":146,"finding":147,"direction":55,"opportunity":148},"评估五种深度卷积神经网络在自然田间条件下对绿豆叶部病害的识别性能。","使用5617张田间图像，五种DCNN架构，迁移学习与微调，数据增强，Grad-C","InceptionV3准确率最高（98.47%），所有模型AUC>0.997，Grad-CAM显示基","可扩展至其他作物多病害识别，或开发轻量级模型用于移动端实时诊断，结合无人机图像实现大范围监测。","智慧农业 \u002F 农业物联网","2026-09-04T23:30:11.523875Z",{"id":152,"title":153,"url":154,"summary":155,"summary_zh":156,"content":9,"source_name":157,"source_url":154,"published_at":158,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":159,"score_detail":160,"sources":163,"tags":165,"search_phrases":168,"slug":171,"view_count":36,"doi":172,"paper":173,"created_at":192},3368,"A PCA-based deep feature optimization framework for explainable orange fruit disease classification","https:\u002F\u002Fdoi.org\u002F10.1186\u002Fs12870-026-09984-8","Accurate classification of orange fruit diseases is important for precision agriculture and yield protection. This study develops and rigorously benchmarks a hybrid deep-feature framework for classifying Black Spot, Canker, Fresh, and Greening oranges (1,090 images), combining deep feature extraction, PCA-based dimensionality reduction, and classical machine-learning classification. Eight backbones (seven CNNs and a Vision Transformer, ViT) and four classifiers (32 configurations in total) were evaluated under 5 × 5 repeated stratified cross-validation, with PCA fitted exclusively on training-fold features in every iteration to eliminate data leakage. The proposed ViT + PCA+SVM configuration achieved the highest mean accuracy, 99.12%±0.71%, significantly outperforming every CNN-based backbone, including DenseNet201 + PCA + SVM (98.48%±0.81%, p \u003C 0.001). A dedicated variance-retention sensitivity analysis justifies the 98% threshold used throughout, and ablation experiments confirm that PCA substantially reduces feature dimensionality (by ~ 55.7% for ViT and ~ 76.6% for DenseNet201) without a significant loss of accuracy for either backbone. Explainability analysis — occlusion sensitivity and SHAP for the proposed ViT model, and Grad-CAM and SHAP for the DenseNet201 comparison model — shows that both configurations base predictions on biologically relevant, disease-affected regions of the fruit rather than spurious cues. These results identify ViT + PCA+SVM as the most accurate configuration evaluated, with DenseNet201 + PCA + SVM as a closely competitive, more compact convolutional alternative for intelligent orchard disease-monitoring systems.","橙类果实病害的准确分类对精准农业和产量保护具有重要意义。本研究开发并严格基准测试了一种混合深度特征框架，用于对黑斑病、溃疡病、新鲜和黄龙病橙类（1，090张图像）进行分类，该框架结合了深度特征提取、基于PCA的降维和经典机器学习分类。在5×5重复分层交叉验证下评估了八种骨干网络（七种CNN和一种视觉Transformer，ViT）和四种分类器（共32种配置），每次迭代中PCA仅在训练折特征上拟合以消除数据泄漏。所提出的ViT + PCA+SVM配置取得了最高平均准确率，为99.12%±0.71%，显著优于所有基于CNN的骨干网络，包括DenseNet201 + PCA + SVM（98.48%±0.81%，p \u003C 0.001）。专门的方差保留敏感性分析证明了全程使用的98%阈值是合理的，消融实验证实PCA大幅降低了特征维度（ViT约降低55.7%，DenseNet201约降低76.6%），且两种骨干网络均无显著准确率损失。可解释性分析——对所提出的ViT模型采用遮挡敏感性和SHAP，对DenseNet201对比模型采用Grad-CAM和SHAP——表明两种配置均基于果实中生物学相关的病害影响区域而非虚假线索进行预测。这些结果确定ViT + PCA+SVM为所评估的最准确配置，而DenseNet201 + PCA + SVM则是一种竞争力接近且更紧凑的卷积替代方案，可用于智能果园病害监测系统。","BMC Plant Biology","2026-09-23T00:00:00Z",79,{"impact":161,"substance":73,"depth":72,"authority":20,"freshness":21,"relevant":22,"comment":162},16,"方法严谨、数据与消融实验充分，对果园智能病害监测有实用参考价值，但属细分技术论文，产业影响有限。",[164],{"name":157,"url":154},[27,28,166,30,167],"可解释AI","柑橘种植",[169,170],"柑橘病害 深度学习 分类","ViT PCA SVM 病害识别","柑橘病害深度学习分类-3368","10.1186\u002Fs12870-026-09984-8",{"doi":172,"openalex_id":174,"authors":175,"venue":157,"cited_by_count":36,"oa_url":154,"card":187,"direction":55,"ingested_from":57},"W7214068709",[176,178,180,182,185],{"name":177,"orcid":9},"Amruta Hingmire",{"name":179,"orcid":9},"Avinash Golande",{"name":181,"orcid":9},"Vinodkumar Bhutnal",{"name":183,"orcid":184},"Sagar Dhanraj Pande","https:\u002F\u002Forcid.org\u002F0000-0003-4506-6997",{"name":186,"orcid":9},"Tanuja Pande",{"tldr":188,"method":189,"finding":190,"direction":55,"opportunity":191},"提出PCA深度特征优化框架，用ViT+SVM分类橙子病害，准确率达99.12%。","8种骨干网络提取特征，PCA降维，4种分类器，5×5交叉验证。","ViT+PCA+SVM最优，PCA降维超55%且精度不降，可解释性验证有效。","可探索轻量化模型在移动端或边缘设备的实时病害检测与多作物泛化。","2026-09-24T23:30:34.116938Z",{"id":194,"title":195,"url":196,"summary":197,"summary_zh":198,"content":9,"source_name":199,"source_url":196,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":200,"score_detail":201,"sources":203,"tags":205,"search_phrases":208,"slug":211,"view_count":36,"doi":212,"paper":213,"created_at":234},3277,"Downscaling of SMAP Soil Moisture Based on the Transformer Algorithm in Anhui Province","https:\u002F\u002Fdoi.org\u002F10.3390\u002Frs18193272","Soil moisture (SM) is critical for climate, water, and agriculture, but Soil Moisture Active Passive (SMAP) passive microwave products have coarse resolution, limiting regional applications. This study develops an SM downscaling framework based on Transformer and its variants (PatchTST and iTransformer), integrating multi-source satellite and groundwater data to generate 1 km daily SM products (2015–2022). Compared with Random Forest (RF), Long Short-Term Memory (LSTM), and Convolutional Neural Network–LSTM (CNN-LSTM), Transformer and its variants achieve superior accuracy and generalization. Validated against in situ measurements and SMCI1.0, the Transformer-downscaled SM product achieved the best accuracy with ubRMSE = 0.0372 m3\u002Fm3 and RMSE = 0.0591 m3\u002Fm3. The downscaled SM dataset not only captured finer spatial details but also preserved the spatial patterns and seasonal dynamics of the original SMAP product and showed good responsiveness to precipitation events. Feature importance analysis revealed that, aside from precipitation, the diurnal land surface temperature difference had a greater impact on SM than individual daytime or nighttime land surface temperature, ranking just below vegetation indices and soil texture factors, while groundwater level showed higher importance than elevation and surface temperature. This study confirms the effectiveness of Transformer-based models for SM spatial downscaling, providing a novel framework integrating remote sensing and deep hydrological information to generate accurate 1 km SM products.","土壤水分（SM）对气候、水资源和农业至关重要，但土壤水分主动被动（SMAP）被动微波产品分辨率较粗，限制了区域应用。本研究构建了一个基于Transformer及其变体（PatchTST和iTransformer）的土壤水分降尺度框架，融合多源卫星和地下水数据，生成1 km日尺度土壤水分产品（2015—2022年）。与随机森林（RF）、长短期记忆网络（LSTM）和卷积神经网络—长短期记忆网络（CNN-LSTM）相比，Transformer及其变体取得了更高的精度和泛化能力。利用站点实测数据和SMCI1.0进行验证，Transformer降尺度土壤水分产品精度最优，ubRMSE = 0.0372 m³\u002Fm³，RMSE = 0.0591 m³\u002Fm³。降尺度土壤水分数据集不仅捕捉到了更精细的空间细节，还保留了原始SMAP产品的空间格局和季节动态，并对降水事件表现出良好的响应。特征重要性分析表明，除降水外，昼夜地表温差对土壤水分的影响大于单独的白天或夜间地表温度，其重要性仅次于植被指数和土壤质地因子，而地下水埋深的重要性高于高程和地表温度。本研究证实了基于Transformer的模型在土壤水分空间降尺度中的有效性，为融合遥感和深层水文信息生成准确的1 km土壤水分产品提供了一种新框架。","Remote Sensing",81,{"impact":72,"substance":73,"depth":72,"authority":20,"freshness":21,"relevant":22,"comment":202},"基于Transformer的SMAP土壤水分1km降尺度研究，方法新颖、验证充分，对区域农业旱情监测有实用价值。",[204],{"name":199,"url":196},[27,28,29,206,207],"遥感","土壤墒情",[209,210],"SMAP 土壤水分 降尺度","Transformer 土壤水分 安徽","SMAP土壤水分降尺度-3277","10.3390\u002Frs18193272",{"doi":212,"openalex_id":214,"authors":215,"venue":199,"cited_by_count":36,"oa_url":196,"card":228,"direction":232,"ingested_from":57},"W7208807695",[216,218,220,222,224,226],{"name":217,"orcid":9},"Yuyang Fan",{"name":219,"orcid":9},"Jianwei Ma",{"name":221,"orcid":9},"Mengmeng Li",{"name":223,"orcid":9},"Changqing Ke",{"name":225,"orcid":9},"Bin Cheng",{"name":227,"orcid":9},"Zheng Duan",{"tldr":229,"method":230,"finding":231,"direction":232,"opportunity":233},"基于Transformer及变体融合多源卫星与地下水数据，将SMAP土壤湿度降尺度至1km日尺度。","Transformer、PatchTST、iTransformer，融合多源卫星","Transformer降尺度产品精度最优（ubRMSE=0.0372），保留原产品时空格局并响应降水","农业遥感与作物表型","可探索Transformer降尺度产品在区域干旱监测、灌溉决策及作物估产中的耦合应用。","2026-09-23T23:30:19.132307Z",{"id":236,"title":237,"url":238,"summary":239,"summary_zh":240,"content":9,"source_name":241,"source_url":238,"published_at":242,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":243,"score_detail":244,"sources":247,"tags":249,"search_phrases":252,"slug":255,"view_count":36,"doi":256,"paper":257,"created_at":290},3165,"Explainable growth stage classification of cacao (Theobroma cacao L.) leaves and key feature visualization using vision transformer and transfer learning","https:\u002F\u002Fdoi.org\u002F10.3389\u002Ffpls.2026.1885906","Accurate and objective plant phenotyping is crucial for optimizing agricultural practices, understanding plant development, and enabling rapid responses to environmental changes. Traditional methods, often relying on visual observation, can be subjective, time-consuming, and may overlook subtle but important differences. This study demonstrates the power of combining digital imaging with deep learning to classify plant material with high accuracy, even when visual differences are minimal. We focused on differentiating between stage D (light green) and stage E (dark green) leaves of cacao ( Theobroma cacao L. ), which are visually very similar in size and overall structure. Using cleared and stained leaves of the SCA 6 genotype to highlight the venation network, we trained a Vision Transformer (ViT) model, a deep learning architecture, on image patches. At the patch level, the model achieved an overall accuracy of 97.03% on an independent test set, with a recall of 96.0% for stage D and 98.3% for stage E. At the whole-leaf level, majority voting correctly classified 14 of 15 independent test leaves (93.3%). Attention maps indicated that image regions containing the midrib and primary lateral veins contributed strongly to classification. These attention maps identify discriminative image regions, but they do not by themselves determine the biological mechanism underlying the signal. The major-vein signal may reflect developmental differences in vein-associated structure, stage-associated differences in Safranin O uptake or optical density, tissue thickness, or a combination of these factors. Because vascular anatomy, lignification, hydraulic conductance, phloem loading, and source–sink status were not directly measured, these mechanisms are treated as hypotheses requiring future anatomical, histochemical, and physiological validation. Thus, this study provides a proof-of-concept for interpretable image-based classification of stage D and stage E leaves within greenhouse-grown SCA 6 cacao. Extension to other cacao genotypes, field-grown plants, independent seasons, staining batches, stress detection, species identification, genotype discrimination, or precision-agriculture deployment will require external validation.","准确、客观的植物表型分析对于优化农业实践、理解植物发育以及快速响应环境变化至关重要。传统方法通常依赖视觉观察，可能具有主观性、耗时，并且可能忽略细微但重要的差异。本研究展示了将数字成像与深度学习相结合，即使在视觉差异极小的情况下，也能以高精度对植物材料进行分类。我们聚焦于区分可可（Theobroma cacao L.）的D期（浅绿色）和E期（深绿色）叶片，这些叶片在大小和整体结构上视觉上非常相似。利用SCA 6基因型的透明染色叶片以突出脉序网络，我们在图像块上训练了Vision Transformer（ViT）模型，一种深度学习架构。在图像块水平上，该模型在独立测试集上达到了97.03%的总体准确率，D期的召回率为96.0%，E期为98.3%。在整叶水平上，多数投票正确分类了15片独立测试叶片中的14片（93.3%）。注意力图表明，包含中脉和初级侧脉的图像区域对分类贡献显著。这些注意力图识别了具有判别力的图像区域，但它们本身并不能确定信号背后的生物学机制。主脉信号可能反映了脉相关结构的发育差异、番红O摄取或光密度的阶段相关差异、组织厚度，或这些因素的组合。由于未直接测量维管解剖结构、木质化、水力导度、韧皮部装载和源–库状态，这些机制被视为假设，需要未来的解剖学、组织化学和生理学验证。因此，本研究为温室种植的SCA 6可可中D期和E期叶片的可解释图像分类提供了概念验证。扩展到其他可可基因型、田间种植植株、独立季节、染色批次、胁迫检测、物种鉴定、基因型区分或精准农业部署将需要外部验证。","Frontiers in Plant Science","2026-09-21T00:00:00Z",71,{"impact":17,"substance":18,"depth":19,"authority":245,"freshness":21,"relevant":22,"comment":246},13,"方法新颖、数据可靠的可解释作物表型概念验证研究，但属实验室小样本，产业影响有限。",[248],{"name":241,"url":238},[27,28,29,250,251],"可可","作物表型",[253,254],"可可 叶片 生长阶段 分类","Vision Transformer 作物表型","可可叶片生长阶段分类-3165","10.3389\u002Ffpls.2026.1885906",{"doi":256,"openalex_id":258,"authors":259,"venue":241,"cited_by_count":36,"oa_url":238,"card":285,"direction":232,"ingested_from":57},"W7213901640",[260,262,265,267,269,271,274,277,280,282],{"name":261,"orcid":9},"Ezekiel Ahn",{"name":263,"orcid":264},"Eun-Sung Park","https:\u002F\u002Forcid.org\u002F0000-0001-6826-2865",{"name":266,"orcid":9},"Moon S. Kim",{"name":268,"orcid":9},"Hangi Kim",{"name":270,"orcid":9},"Lalit M. Kandpal",{"name":272,"orcid":273},"Sunchung Park","https:\u002F\u002Forcid.org\u002F0000-0002-7398-9476",{"name":275,"orcid":276},"Seunghyun Lim","https:\u002F\u002Forcid.org\u002F0000-0003-3023-4863",{"name":278,"orcid":279},"Lyndel W. Meinhardt","https:\u002F\u002Forcid.org\u002F0000-0001-8299-2629",{"name":281,"orcid":9},"Byoung-Kwan Cho",{"name":283,"orcid":284},"Insuck Baek","https:\u002F\u002Forcid.org\u002F0000-0003-1044-349X",{"tldr":286,"method":287,"finding":288,"direction":232,"opportunity":289},"用ViT和迁移学习对可可叶D、E期进行可解释分类，准确率达97%。","透明染色叶片图像块训练ViT，注意力图可视化关键区域。","模型准确区分D\u002FE期，中脉和主侧脉区域贡献最大。","可扩展到多基因型、田间、胁迫检测，并验证脉信号生物学机制。","2026-09-22T23:30:19.972949Z",{"id":292,"title":293,"url":294,"summary":295,"summary_zh":296,"content":9,"source_name":297,"source_url":294,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":298,"score_detail":299,"sources":301,"tags":303,"search_phrases":305,"slug":308,"view_count":36,"doi":309,"paper":310,"created_at":336},3153,"Explainable optimized deep learning and generative AI based framework for finger millet disease detection in smart agriculture","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs44163-026-02222-y","Abstract Eleusine coracana, locally known as finger millet (ragi), is a wholesome, climate-resilient crop. It is a staple food in semi-dry and dry areas of Asia and Africa. The productivity of this crop can be severely reduced by diseases such as downy, mottle, seedling, smut, and wilt. Conventional diagnosis methods are often time-consuming and unsuitable for large-scale farming. Therefore,Deep Learning (DL) offers an alternative for smart commercial farming that can streamline disease screening and improve productivity. This work offers an AI-integrated framework for early disease intervention combining DL, Explainable AI (XAI), and an image-based decision-support interface. The finger millet (ragi) dataset from Kaggle was used to train and evaluate custom CNN, VGG16, and ResNet50 models. To reduce feature redundancy and improve generalization, Grey Wolf Optimizer (GWO) based feature selection was applied to the penultimate-layer features of each model. Custom CNN, VGG16, and ResNet50 achieved accuracies of 89.40%, 83.11%, and 91.83%, respectively. Further, the use of GWO improved the accuracies to 97.71%, 90.84%, and 98.36%, respectively. ResNet50 achieved the highest accuracy 98.36% and was selected as the final backbone. The final ResNet50 model is integrated with the Gemini API to provide disease-specific recommendations based on the predicted class and user queries. To increase prediction transparency, Grad-CAM was used to highlight image regions that influenced the model output. The integration of DL, XAI, and generative AI provides a scalable approach for detecting and managing finger millet diseases. The proposed work boosts precision agriculture by providing AI-based and XAI-supported decision-making, while fostering sustainable agricultural practices.","摘要 穇子（Eleusine coracana），当地称为指黍（ragi），是一种有益健康且气候适应性强的作物。它是亚洲和非洲半干旱及干旱地区的主食。霜霉病、斑驳病、苗枯病、黑穗病和枯萎病等病害可严重降低该作物的产量。传统诊断方法往往耗时且不适合大规模种植。因此，深度学习（DL）为智能商业农业提供了一种替代方案，可简化病害筛查并提高生产力。本研究提出了一种人工智能集成框架，用于早期病害干预，结合了深度学习、可解释人工智能（XAI）和基于图像的决策支持界面。使用来自Kaggle的指黍（ragi）数据集训练和评估了自定义CNN、VGG16和ResNet50模型。为减少特征冗余并提高泛化能力，将灰狼优化器（GWO）基于特征选择应用于每个模型的倒数第二层特征。自定义CNN、VGG16和ResNet50分别达到了89.40%、83.11%和91.83%的准确率。此外，使用GWO将准确率分别提高到97.71%、90.84%和98.36%。ResNet50达到了最高准确率98.36%，并被选为最终骨干网络。最终的ResNet50模型与Gemini API集成，根据预测类别和用户查询提供针对特定病害的建议。为提高预测透明度，使用Grad-CAM突出显示影响模型输出的图像区域。深度学习、可解释人工智能和生成式人工智能的集成提供了一种可扩展的方法，用于检测和管理指黍病害。所提出的工作通过提供基于人工智能和可解释人工智能支持的决策，促进了精准农业，同时推动了可持续农业实践。","Discover Artificial Intelligence",75,{"impact":114,"substance":18,"depth":72,"authority":17,"freshness":13,"relevant":22,"comment":300},"将深度学习、可解释AI与生成式AI结合用于指状粟病害识别，方法新颖、数据充分，对智慧农业有参考价值。",[302],{"name":297,"url":294},[27,28,166,30,304],"小米作物",[306,307],"ResNet50 GWO 病害识别","农业人工智能 小米作物 智慧农业 病害识别","ResNet50GWO病害识别-3153","10.1007\u002Fs44163-026-02222-y",{"doi":309,"openalex_id":311,"authors":312,"venue":297,"cited_by_count":36,"oa_url":294,"card":331,"direction":149,"ingested_from":57},"W7213978567",[313,316,318,320,322,325,328],{"name":314,"orcid":315},"Sunil Kumar Mohapatra","https:\u002F\u002Forcid.org\u002F0000-0002-5865-095X",{"name":317,"orcid":9},"A. Sanjib Kumar Patro",{"name":319,"orcid":9},"Lulen Kumar Sahu",{"name":321,"orcid":9},"Chinmaye Dora",{"name":323,"orcid":324},"Sujata Chakravarty","https:\u002F\u002Forcid.org\u002F0000-0002-1293-5378",{"name":326,"orcid":327},"Kshira Sagar Sahoo","https:\u002F\u002Forcid.org\u002F0000-0002-6435-5738",{"name":329,"orcid":330},"Byomakesh Mahapatra","https:\u002F\u002Forcid.org\u002F0000-0002-9126-1729",{"tldr":332,"method":333,"finding":334,"direction":55,"opportunity":335},"提出融合深度学习、可解释AI与生成式AI的指状粟病害检测框架，实现高精度识别与决策支持。","用Kaggle指状粟图像训练CNN、VGG16、ResNet50，并用灰狼优化做","GWO特征选择显著提升精度，ResNet50达98.36%，结合Grad-CAM与Gemini AP","可探索轻量化模型与多作物泛化，并将XAI与生成式建议在田间移动端实时验证。","2026-09-22T23:30:10.255790Z"]