[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-4044":3,"related-4044":78},{"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,"primary_topic":9,"secondary_topics":32,"classification":33,"search_phrases":57,"slug":60,"view_count":41,"doi":61,"paper":62,"created_at":77},4044,"BeLFG-Net: A Lightweight Betel Leaf Freshness Grading Network via Hybrid Depthwise Separable Convolutions","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.atech.2026.102597","Post-harvest quality grading of betel leaves ( Piper betle L.) is critical to supply chain efficiency across South and Southeast Asia, yet assessment remains an entirely manual process susceptible to inter-rater inconsistency and throughput bottlenecks. Visual freshness cues—saturated green pigmentation, surface desiccation, and brown discoloration—form a continuous rather than discrete progression, making threshold-based color analysis insufficient. We propose BeLFG-Net, a Betel Leaf Freshness Grading Network, that addresses this challenge through a hybrid four-block encoder combining standard convolutions for initial feature establishment with depthwise separable pairs for parameter-efficient cross-channel mixing, followed by global average pooling and a compact classification head. The key efficiency gain arises from the depthwise separable factorization, which decouples spatial filtering from channel-wise integration and reduces computation approximately 8–9 × relative to equivalent standard convolutions, yielding a total of only 979,651 parameters. We further introduce a progressive dropout schedule ( p ∈ {0.25, 0.25, 0.25, 0.50}) that imposes strongest regularization at the bottleneck transition, effectively curbing overfitting on the 766-image field dataset augmented to 6,262 samples. On the three-class freshness discrimination task—fresh, moderate, and stale—BeLFG-Net achieves 95.72% test accuracy with per-class F1-scores of 0.96, 0.96, and 0.95, outperforming ResNet-50 (93.77%) and MobileNetV2 (91.65%) while using 26 × and 3.5 × fewer parameters, respectively. Ablation experiments confirm that the hybrid depthwise separable block design, progressive dropout, and global average pooling each contribute independently to generalization, with combined improvements of + 6.76 % over a standard CNN baseline. These results establish that domain-specific lightweight architectures can deliver high-accuracy post-harvest classification under the resource constraints common to agricultural edge deployments.","采后槟榔叶（Piper betle L.）品质分级对南亚和东南亚供应链效率至关重要，然而目前评估仍完全依赖人工，易受评估者间不一致和吞吐瓶颈的影响。视觉新鲜度线索——饱和绿色色素、表面干燥和褐色变色——呈连续而非离散变化，使得基于阈值的颜色分析不足以应对。我们提出BeLFG-Net，即槟榔叶新鲜度分级网络（Betel Leaf Freshness Grading Network），通过混合四模块编码器应对这一挑战，该编码器将用于初始特征建立的标准卷积与用于参数高效跨通道混合的深度可分离卷积对相结合，随后采用全局平均池化和紧凑分类头。关键效率提升源于深度可分离分解，它将空间滤波与通道整合解耦，相较于等效标准卷积减少约8–9倍计算量，总参数量仅为979,651。我们进一步引入渐进式dropout调度（p ∈ {0.25, 0.25, 0.25, 0.50}），在瓶颈过渡处施加最强正则化，有效抑制了在766张田间图像数据集（增强至6,262个样本）上的过拟合。在三类新鲜度判别任务——新鲜、中等和不新鲜——中，BeLFG-Net达到95.72%的测试准确率，各类别F1分数分别为0.96、0.96和0.95，优于ResNet-50（93.77%）和MobileNetV2（91.65%），同时参数量分别减少26倍和3.5倍。消融实验证实，混合深度可分离模块设计、渐进式dropout和全局平均池化各自独立促进泛化，相较标准CNN基线合计提升+6.76%。这些结果表明，领域特定的轻量级架构能够在农业边缘部署常见的资源约束下实现高精度采后分类。",null,"Smart Agricultural Technology","2026-10-01T00:00:00Z","论文",10,false,74,{"impact":17,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":23},12,22,18,13,9,1,"面向蒌叶采后分级的轻量化深度网络研究，方法新颖、数据与对比实验扎实，对农业边缘部署有参考价值，但属细分作物应用，产业影响有限。",[25],{"name":10,"url":6},[27,28,29,30,31],"智慧农业","农业人工智能","采后处理","农产品分级","轻量化模型",[],{"status":34,"locked":14,"version":35,"method":36,"evidence":37},"unclassified","2026-10-22.2","rules",{"candidates":38,"agricultural_evidence":55,"reason":56},[39,45,50],{"code":40,"score":41,"supported":14,"title":42,"abstract":43,"tags":44},"I2",2,[],[],[28,31],{"code":46,"score":22,"supported":14,"title":47,"abstract":48,"tags":49},"E1",[],[],[30],{"code":51,"score":22,"supported":14,"title":52,"abstract":53,"tags":54},"I1",[],[],[27],true,"缺少足够的主题证据",[58,59],"蒌叶 新鲜度 分级","BeLFG-Net 轻量网络","蒌叶新鲜度分级-4044","10.1016\u002Fj.atech.2026.102597",{"doi":61,"openalex_id":63,"authors":64,"venue":10,"cited_by_count":69,"oa_url":6,"card":70,"direction":74,"ingested_from":76},"W7216219039",[65,67],{"name":66,"orcid":9},"Fahim Ur Rahman",{"name":68,"orcid":9},"Afraz Ul Haque",0,{"tldr":71,"method":72,"finding":73,"direction":74,"opportunity":75},"提出轻量网络BeLFG-Net，实现槟榔叶采后新鲜度三级自动分级。","混合深度可分离卷积编码器+渐进Dropout，766张图像增强至6262张。","测试准确率95.72%，优于ResNet-50和MobileNetV2且参数量大幅减少。","农业人工智能与决策模型","可迁移至其他叶菜\u002F农产品采后分级，并探索边缘设备实时部署与多模态融合。","openalex","2026-10-03T23:30:03.701027Z",{"total":79,"page":22,"page_size":79,"items":80},6,[81,153,210,264,339,383],{"id":82,"title":83,"url":84,"summary":85,"summary_zh":86,"content":9,"source_name":10,"source_url":84,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":87,"score_detail":88,"sources":91,"tags":93,"primary_topic":96,"secondary_topics":102,"classification":103,"search_phrases":122,"slug":125,"view_count":126,"doi":127,"paper":128,"created_at":152},4163,"Severity Grading of Early Visible Symptoms of Sugarcane Pokkah Boeng Disease Based on Multi-Scale Dynamic Feature Fusion and Channel-Spatial Interaction: An On-Device Framework","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.atech.2026.102607","Sugarcane pokkah boeng disease is a fungal disease caused by pathogens in the Fusarium fujikuroi species complex (FFSC). Early visible symptoms include chlorosis and yellowing at the base of young leaves, leaf wrinkling, and morphological abnormalities, which may differ only subtly from healthy leaves and make manual inspection and severity assessment difficult. Existing studies have paid limited attention to severity grading of these early visible symptoms, while some models remain too complex for mobile deployment. To address these issues, this study established a four-grade severity standard using leaf colour alteration, leaf morphological abnormalities, and visible abnormalities in the growing-point region as visual symptom anchors, and constructed an image dataset for sugarcane pokkah boeng disease. A lightweight grading framework based on MobileNetV2 was then developed for this task, incorporating multi-scale dynamic feature fusion with a fast channel-spatial interaction module to enhance multi-scale and channel-spatial representations of severity-related visual features. The proposed method achieved Accuracy, Precision, Recall, and Macro-F1 of 96.29% ± 0.73%, 95.32% ± 0.79%, 95.99% ± 0.78%, and 95.43% ± 0.77%, respectively, outperforming the main comparison models overall. On-device accuracy was 96.53% under both FP32 and INT8 precision. After INT8 quantisation, the ONNX model size decreased from 8.63 MB to 2.57 MB, and single-image inference latency from 30.819 ± 4.625 ms to 10.794 ± 0.999 ms. This study provides a lightweight solution for image-based visual severity grading and mobile-assisted assessment of early visible symptoms of sugarcane pokkah boeng disease, thereby supporting technical support for mobile-assisted severity assessment and disease management.","甘蔗凤梨病（pokkah boeng disease）是由藤仓镰刀菌复合种（Fusarium fujikuroi species complex，FFSC）病原菌引起的真菌病害。早期可见症状包括幼叶基部褪绿黄化、叶片皱缩及形态异常，这些症状与健康叶片差异细微，导致人工巡检和严重度评估困难。现有研究对上述早期可见症状的严重度分级关注有限，且部分模型过于复杂，难以在移动端部署。针对上述问题，本研究以叶片颜色改变、叶片形态异常及生长点区域可见异常作为视觉症状锚点，建立了四级严重度标准，并构建了甘蔗凤梨病图像数据集。随后，基于MobileNetV2开发了适用于该任务的轻量化分级框架，融合多尺度动态特征融合与快速通道-空间交互模块，以增强严重度相关视觉特征的多尺度和通道-空间表征能力。所提方法在准确率、精确率、召回率和Macro-F1上分别达到96.29% ± 0.73%、95.32% ± 0.79%、95.99% ± 0.78%和95.43% ± 0.77%，总体优于主要对比模型。在FP32和INT8精度下，端侧准确率均为96.53%。经INT8量化后，ONNX模型大小从8.63 MB降至2.57 MB，单张图像推理延迟从30.819 ± 4.625 ms降至10.794 ± 0.999 ms。本研究为甘蔗凤梨病早期可见症状的基于图像的视觉严重度分级与移动辅助评估提供了轻量化解决方案，从而为移动辅助严重度评估和病害管理提供技术支持。",79,{"impact":19,"substance":18,"depth":89,"authority":20,"freshness":21,"relevant":22,"comment":90},17,"面向甘蔗凤梨病早期症状的轻量化端侧分级框架，方法新颖、指标扎实，具备移动辅助防控的实用价值。",[92],{"name":10,"url":84},[27,28,31,94,95],"甘蔗病害","病害分级",{"code":97,"slug":98,"name":99,"path":100,"parent_code":101},"A6","cash-crops","棉麻糖与特色经作","\u002Ftopic\u002Ffield-crops\u002Fcash-crops","A",[],{"status":104,"locked":14,"version":35,"method":36,"evidence":105},"classified",{"candidates":106,"agricultural_evidence":55},[107,114,118],{"code":97,"score":108,"supported":55,"title":109,"abstract":111,"tags":113},14,[110],"sugarcane",[112,110],"甘蔗",[],{"code":40,"score":41,"supported":14,"title":115,"abstract":116,"tags":117},[],[],[28,31],{"code":51,"score":22,"supported":14,"title":119,"abstract":120,"tags":121},[],[],[27],[123,124],"甘蔗 凤梨病 分级","MobileNetV2 病害识别","甘蔗凤梨病分级-4163",3,"10.1016\u002Fj.atech.2026.102607",{"doi":127,"openalex_id":129,"authors":130,"venue":10,"cited_by_count":69,"oa_url":84,"card":147,"direction":74,"ingested_from":76},"W7218607722",[131,134,136,138,140,142,144],{"name":132,"orcid":133},"Cuimin Sun","https:\u002F\u002Forcid.org\u002F0000-0003-4174-1094",{"name":135,"orcid":9},"Xiaojie Qin",{"name":137,"orcid":9},"Junyang Dou",{"name":139,"orcid":9},"Huijing Chen",{"name":141,"orcid":9},"Yongping Liao",{"name":143,"orcid":9},"Songjie Yuan",{"name":145,"orcid":146},"Chengwu Zou","https:\u002F\u002Forcid.org\u002F0000-0002-2836-2930",{"tldr":148,"method":149,"finding":150,"direction":74,"opportunity":151},"提出轻量级框架，对甘蔗凤梨病早期可见症状进行四级严重度分级。","基于MobileNetV2，融合多尺度动态特征与通道-空间交互模块，构建图像数据","准确率96.29%，INT8量化后模型仅2.57MB，推理延迟10.79ms，适合移动端部署。","可扩展至其他作物早期病害分级，并探索田间复杂光照下的鲁棒性与多病害联合识别。","2026-10-04T23:30:03.798689Z",{"id":154,"title":155,"url":156,"summary":157,"summary_zh":9,"content":9,"source_name":158,"source_url":9,"published_at":11,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":159,"score_detail":160,"sources":163,"tags":165,"primary_topic":168,"secondary_topics":174,"classification":175,"search_phrases":195,"slug":198,"view_count":199,"doi":9,"paper":200,"created_at":209},3962,"华中农业大学柑橘类果实泛化识别与定位研究——YOLOv8n-Light轻量化部署","https:\u002F\u002Fnews.foodmate.net\u002F2026\u002F10\u002Fwap_754150.html","华中农业大学工学院马京奥、刘洁副教授团队研究YOLOv8n-Light模型在柑橘类果实识别中的应用，模型可同时识别橙、橘、柚三种果实，计算量较原始YOLOv8降低58%，工作站CPU推理速度提升45.9%（达15.35FPS），树莓派4B推理速度提升38.2%（达0.76FPS）；多轴机械臂采摘作业控制中，果实中心坐标X、Y、Z方向平均误差均控制在10mm内，半径误差小于6mm。312次果园采摘测试验证高兼容性与实用性。研究得到国家重点研发计划、国家柑橘产业技术体系资助。","食品伙伴网·2026年10月01日",82,{"impact":19,"substance":18,"depth":161,"authority":20,"freshness":13,"relevant":22,"comment":162},19,"高校团队提出轻量化柑橘识别模型并完成果园实测，方法新颖、数据扎实，对智能采摘装备有参考价值。",[164],{"name":158,"url":156},[27,28,166,167,31],"采摘机器人","柑橘",{"code":169,"slug":170,"name":171,"path":172,"parent_code":173},"B2","fruit","果树","\u002Ftopic\u002Fhorticulture\u002Ffruit","B",[],{"status":104,"locked":14,"version":35,"method":36,"evidence":176},{"candidates":177,"agricultural_evidence":55},[178,182,186,191],{"code":169,"score":17,"supported":55,"title":179,"abstract":180,"tags":181},[167],[167],[167],{"code":40,"score":41,"supported":14,"title":183,"abstract":184,"tags":185},[],[],[28,31],{"code":187,"score":22,"supported":14,"title":188,"abstract":189,"tags":190},"F6",[],[],[166],{"code":51,"score":22,"supported":14,"title":192,"abstract":193,"tags":194},[],[],[27],[196,197],"华中农业大学 柑橘 采摘机器人","YOLOv8n-Light 柑橘识别","华中农业大学柑橘采摘机器人-3962",5,{"doi":9,"openalex_id":9,"authors":201,"venue":9,"cited_by_count":69,"oa_url":9,"card":202,"direction":206,"ingested_from":208},[],{"tldr":203,"method":204,"finding":205,"direction":206,"opportunity":207},"提出YOLOv8n-Light轻量化模型，实现橙橘柚三种柑橘果实泛化识别与定位，并部署于树莓派和采摘","改进YOLOv8n轻量化模型，工作站与树莓派4B部署，多轴机械臂采摘测试312次","计算量降58%，工作站CPU推理15.35FPS，果实中心坐标误差\u003C10mm，半径误差\u003C6mm。","智慧农业 \u002F 农业物联网","可探索多品种泛化模型在边缘设备上的精度-速度权衡，及跨季节、遮挡场景下的鲁棒性提升。","agent","2026-10-02T11:53:16.558461Z",{"id":211,"title":212,"url":213,"summary":214,"summary_zh":215,"content":9,"source_name":216,"source_url":213,"published_at":217,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":218,"score_detail":219,"sources":222,"tags":224,"primary_topic":227,"secondary_topics":232,"classification":233,"search_phrases":245,"slug":248,"view_count":126,"doi":249,"paper":250,"created_at":263},3517,"A Resource-Efficient Hybrid CNN-LSTM Network for Image-Based Bean Leaf Disease Classification","https:\u002F\u002Fdoi.org\u002F10.3390\u002Fjimaging12100468","Accurate and resource-efficient automated diagnosis is a cornerstone of modern agricultural expert systems. While Convolutional Neural Networks (CNNs) have established benchmarks in plant pathology, their ability to capture long-range spatial dependencies is often limited by standard pooling layers, and their high memory footprint hinders deployment on portable devices. This paper proposes a lightweight hybrid CNN-LSTM system for bean leaf disease classification. By integrating an LSTM layer to model the spatial–sequential relationships within feature maps, our hybrid architecture achieves a 94.36% accuracy and 94.38% F1 score while maintaining an exceptionally small footprint of 1.86 MB, a 70% reduction in size compared to traditional CNN-based systems. Furthermore, we provide a systematic evaluation of image augmentation strategies, demonstrating that tailored transformations are superior to generic combinations for maintaining the integrity of diagnostic patterns. Results on the ibean dataset confirm that the proposed system achieves state-of-the-art F1 scores of 99.22% with EfficientNet-B7+LSTM, providing a potentially robust and scalable framework for real-time agricultural decision support in resource-constrained environments. The code and augmented datasets used in this study are publicly available on this GitHub repo.","准确且资源高效的自动化诊断是现代农业专家系统的基石。尽管卷积神经网络（CNN）在植物病理学领域已确立了基准，但其捕捉长程空间依赖关系的能力常受限于标准池化层，且高内存占用阻碍了其在便携设备上的部署。本文提出了一种用于豆叶病害分类的轻量级混合CNN-LSTM系统。通过集成LSTM层来建模特征图内的空间-序列关系，我们的混合架构达到了94.36%的准确率和94.38%的F1分数，同时保持了仅1.86 MB的极小占用，相较于传统基于CNN的系统体积减少了70%。此外，我们系统评估了图像增强策略，表明定制化变换在保持诊断模式完整性方面优于通用组合。在ibean数据集上的结果证实，所提出的系统结合EfficientNet-B7+LSTM达到了99.22%的最先进F1分数，为资源受限环境中的实时农业决策支持提供了一个潜在稳健且可扩展的框架。本研究使用的代码和增强数据集已在此GitHub仓库公开。","Journal of Imaging","2026-09-24T00:00:00Z",78,{"impact":220,"substance":18,"depth":19,"authority":20,"freshness":21,"relevant":22,"comment":221},16,"轻量级CNN-LSTM豆叶病害识别研究，方法新颖、数据可靠且代码开源，对资源受限场景下的农业智能诊断有实用价值。",[223],{"name":216,"url":213},[27,28,225,31,226],"病害识别","豆类作物",{"code":40,"slug":228,"name":229,"path":230,"parent_code":231},"agricultural-ai","农业AI与大模型","\u002Ftopic\u002Fdigital-agriculture\u002Fagricultural-ai","I",[],{"status":104,"locked":14,"version":35,"method":36,"evidence":234},{"candidates":235,"agricultural_evidence":55},[236,241],{"code":40,"score":79,"supported":55,"title":237,"abstract":238,"tags":240},[],[239],"decision support",[28,225,31],{"code":51,"score":22,"supported":14,"title":242,"abstract":243,"tags":244},[],[],[27],[246,247],"ibean dataset 豆叶病害","农业人工智能 轻量化模型 智慧农业 病害识别","ibeandataset豆叶病害-3517","10.3390\u002Fjimaging12100468",{"doi":249,"openalex_id":251,"authors":252,"venue":216,"cited_by_count":69,"oa_url":213,"card":258,"direction":74,"ingested_from":76},"W7154572440",[253,255],{"name":254,"orcid":9},"Hye Jin Rhee",{"name":256,"orcid":257},"Joseph Damilola Akinyemi","https:\u002F\u002Forcid.org\u002F0000-0003-3121-4231",{"tldr":259,"method":260,"finding":261,"direction":74,"opportunity":262},"提出轻量级CNN-LSTM混合网络，用于豆叶病害分类，兼顾高精度与低资源占用。","CNN提取特征后接LSTM建模空间序列关系，在ibean数据集上评估并系统比较图","模型准确率94.36%、F1 94.38%，仅1.86MB，比传统CNN缩小70%，Efficien","可探索面向移动端\u002F边缘设备的超轻量病害诊断模型，并研究增强策略与模型结构的自适应协同优化。","2026-09-25T23:30:59.276381Z",{"id":265,"title":266,"url":267,"summary":268,"summary_zh":269,"content":9,"source_name":270,"source_url":267,"published_at":271,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":272,"score_detail":273,"sources":276,"tags":278,"primary_topic":281,"secondary_topics":282,"classification":283,"search_phrases":303,"slug":306,"view_count":126,"doi":307,"paper":308,"created_at":338},3134,"DAC-Grad-CAM: Defect-aware explainable AI for automated mango quality classification","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.compag.2026.112443","Deep learning models achieve high fruit classification accuracy but lack interpretability; standard Grad-CAM produces diffuse attention without defect-specific localization. Existing explainable AI (XAI) methods for fruit grading rely on qualitative visual inspection, offering no quantitative validation of defect alignment, thereby limiting trust and regulatory acceptance in commercial quality control. This study presents an explainable mango classification framework that integrates classification performance with quantitative interpretability analysis through Defect-Aware Contrastive Grad-CAM (DAC-Grad-CAM). A new method that multiplicatively fuses gradient-based attention with an unsupervised LAB color-space bruise prior. A two-stage pipeline combining YOLO-11n detection with four classification architectures (ResNet18, MobileNetV3, VGG16, and a custom CNN). Evaluated on a three-class mango dataset (premium, class 1, class 2). Transfer-learning architectures substantially outperformed the custom CNN (ResNet18: 97.8%, VGG16: 97.6%, MobileNetV3: 96.5% vs 75.4%), a 22–23 percentage-point gap, with premium-class precision > 0.995. DAC-Grad-CAM produced the most concentrated explanations (entropy 10.02 versus 10.27 for Grad-CAM and 10.18 for SHAP). Validated against 50 manually annotated pixel-level bruise masks, it improved peak localization over standard Grad-CAM (pointing-game accuracy 0.561 versus 0.439) and, on the transfer-learning backbones, exceeded SHAP (0.659 versus 0.615) while requiring 34 ms per region against 4220 ms, a 124-fold reduction in cost. The Defect Focus Score (DFS) enabled task-specific validation, and the agreement between gradient and perturbation-based explanations (r = 0.68–0.75, p \u003C 0.001) confirmed that the localized features are task-genuine rather than method-specific artifacts. Under controlled perturbations emulating varied lighting, resolution, defocus, compression, and occlusion, defect-focused attention remained stable while classification accuracy degraded, and end-to-end throughput reached 17.4 fruit s⁻ 1 . The framework supports a confidence-based human-AI workflow (margin > 0.95, DFS > 0.37, r > 0.60) for trustworthy automated quality control in post-harvest operations.","深度学习模型虽能实现高精度水果分类，却缺乏可解释性；标准Grad-CAM产生的注意力分布弥散，无法实现缺陷特异性定位。现有用于水果分级的可解释人工智能（XAI）方法依赖定性视觉检查，无法对缺陷对齐进行定量验证，从而限制了其在商业质量控制中的可信度与法规接受度。本研究提出了一种可解释芒果分类框架，通过缺陷感知对比Grad-CAM（DAC-Grad-CAM）将分类性能与定量可解释性分析相结合。该方法将基于梯度的注意力与无监督LAB色彩空间瘀伤先验进行乘法融合。采用两阶段流水线，将YOLO-11n检测与四种分类架构（ResNet18、MobileNetV3、VGG16及自定义CNN）相结合。在三类芒果数据集（特级、一级、二级）上进行评估。迁移学习架构显著优于自定义CNN（ResNet18：97.8%，VGG16：97.6%，MobileNetV3：96.5% vs 75.4%），差距达22–23个百分点，特级类精确率> 0.995。DAC-Grad-CAM产生了最集中的解释（熵10.02，对比Grad-CAM的10.27和SHAP的10.18）。针对50个手动标注的像素级瘀伤掩膜进行验证，其峰值定位优于标准Grad-CAM（指向游戏准确率0.561 vs 0.439），且在迁移学习骨干网络上超过SHAP（0.659 vs 0.615），同时每区域仅需34 ms，而SHAP需4220 ms，成本降低124倍。缺陷聚焦分数（DFS）实现了任务特异性验证，基于梯度与基于扰动的解释之间的一致性（r = 0.68–0.75，p \u003C 0.001）证实了所定位特征是任务真实的，而非方法特异性伪影。在模拟不同光照、分辨率、散焦、压缩和遮挡的受控扰动下，缺陷聚焦注意力保持稳定，而分类准确率下降，端到端吞吐量达到17.4果·秒⁻¹。该框架支持基于置信度的人机协作工作流（margin > 0.95，DFS > 0.37，r > 0.60），可用于采后作业中可信赖的自动化质量控制。","Computers and Electronics in Agriculture","2026-09-21T00:00:00Z",81,{"impact":220,"substance":274,"depth":161,"authority":108,"freshness":21,"relevant":22,"comment":275},23,"方法新颖、量化验证扎实的芒果品质分级可解释AI研究，对采后智能质检有直接参考价值。",[277],{"name":270,"url":267},[27,28,279,29,280],"可解释AI","水果分级",{"code":40,"slug":228,"name":229,"path":230,"parent_code":231},[],{"status":104,"locked":14,"version":35,"method":36,"evidence":284},{"candidates":285,"agricultural_evidence":55},[286,294,299],{"code":40,"score":287,"supported":55,"title":288,"abstract":289,"tags":293},11,[],[290,291,292],"深度学习","deep learning","YOLO",[28,279],{"code":169,"score":126,"supported":55,"title":295,"abstract":296,"tags":298},[],[297],"芒果",[],{"code":51,"score":22,"supported":14,"title":300,"abstract":301,"tags":302},[],[],[27],[304,305],"芒果 品质分级 深度学习","DAC-Grad-CAM 芒果 缺陷检测","芒果品质分级深度学习-3134","10.1016\u002Fj.compag.2026.112443",{"doi":307,"openalex_id":309,"authors":310,"venue":270,"cited_by_count":69,"oa_url":267,"card":333,"direction":74,"ingested_from":76},"W7213923596",[311,313,315,317,319,321,324,326,329,331],{"name":312,"orcid":9},"Arshed Ahmed",{"name":314,"orcid":9},"Zhao Zhang",{"name":316,"orcid":9},"Ming Li",{"name":318,"orcid":9},"Shahram Hamza Manzoor",{"name":320,"orcid":9},"Zhanjiang Zhu",{"name":322,"orcid":323},"Rashed Ahmed","https:\u002F\u002Forcid.org\u002F0009-0000-7170-2200",{"name":325,"orcid":9},"Noor Gul",{"name":327,"orcid":328},"Mustafa Mhamed","https:\u002F\u002Forcid.org\u002F0000-0002-3106-669X",{"name":330,"orcid":9},"Bing Liu",{"name":332,"orcid":9},"Mahmoud A. Abdelhamid",{"tldr":334,"method":335,"finding":336,"direction":74,"opportunity":337},"提出DAC-Grad-CAM可解释框架，实现芒果缺陷感知分类与定量验证。","融合梯度注意力与LAB色空间瘀伤先验，YOLO-11n检测加四种分类网络。","ResNet18达97.8%精度，定位优于Grad-CAM与SHAP且快124倍。","可将缺陷感知XAI扩展至多水果多缺陷，并构建人机协同质控标准。","2026-09-22T23:30:02.407104Z",{"id":340,"title":341,"url":342,"summary":343,"summary_zh":9,"content":9,"source_name":344,"source_url":9,"published_at":345,"category":12,"cover_url":9,"hotness":13,"is_selected":14,"score":87,"score_detail":346,"sources":348,"tags":350,"primary_topic":353,"secondary_topics":354,"classification":355,"search_phrases":371,"slug":374,"view_count":79,"doi":9,"paper":375,"created_at":382},3048,"基于双路径注意力与多尺度融合的作物病虫害识别网络DPMFNet","https:\u002F\u002Fwww.mdpi.com\u002F1099-4300\u002F28\u002F9\u002F1032","盐城工学院Hong Zhang、Fagen Song等联合江苏开放大学提出DPMFNet轻量级双路径网络，集成空间-通道双注意力（SCDA）与多尺度深度可分离卷积（MDSC）模块，构建AttMDSCBlock残差结构。在PlantVillage与AI Challenger 2018数据集上DPMFNet仅14.24M参数和2.55G FLOPs，跨注意力机制融合局部细节与全局上下文，轻量化金字塔策略自适应整合多分辨率特征，在复杂农田场景下兼顾精度与可部署性，为嵌入式田间设备提供高性价比方案。","MDPI Entropy 28(9):1032","2026-09-20T00:00:00Z",{"impact":220,"substance":18,"depth":19,"authority":20,"freshness":13,"relevant":22,"comment":347},"轻量级双路径注意力网络在两大公开数据集上兼顾精度与可部署性，对嵌入式田间设备落地有实质参考价值。",[349],{"name":344,"url":342},[27,28,351,352,31],"病虫害识别","作物监测",{"code":40,"slug":228,"name":229,"path":230,"parent_code":231},[],{"status":104,"locked":14,"version":35,"method":36,"evidence":356},{"candidates":357,"agricultural_evidence":55},[358,362,366],{"code":40,"score":287,"supported":55,"title":359,"abstract":360,"tags":361},[351],[],[28,351,31],{"code":51,"score":22,"supported":14,"title":363,"abstract":364,"tags":365},[],[],[27],{"code":367,"score":22,"supported":14,"title":368,"abstract":369,"tags":370},"I3",[],[],[352],[372,373],"盐城工学院 DPMFNet 病虫害识别","PlantVillage AI Challenger 作物病害","盐城工学院DPMFNet病虫害识别-3048",{"doi":9,"openalex_id":9,"authors":376,"venue":9,"cited_by_count":69,"oa_url":9,"card":377,"direction":74,"ingested_from":208},[],{"tldr":378,"method":379,"finding":380,"direction":74,"opportunity":381},"提出轻量级双路径网络DPMFNet，实现复杂农田场景下的作物病虫害高精度识别。","空间-通道双注意力与多尺度深度可分离卷积，构建AttMDSCBlock残差结构。","仅14.24M参数、2.55G FLOPs，在PlantVillage与AI Challenger ","可探索真实田间多病虫害并发与边缘设备实时推理的轻量化自适应识别研究。","2026-09-21T00:04:39.305757Z",{"id":384,"title":385,"url":386,"summary":387,"summary_zh":9,"content":9,"source_name":388,"source_url":9,"published_at":389,"category":12,"cover_url":9,"hotness":390,"is_selected":14,"score":391,"score_detail":392,"sources":394,"tags":399,"primary_topic":402,"secondary_topics":403,"classification":404,"search_phrases":421,"slug":424,"view_count":126,"doi":9,"paper":425,"created_at":432},3046,"基于Flor-YOLO的香石竹鲜切花分级轻量化检测方法","https:\u002F\u002Ffinance.sina.com.cn\u002Froll\u002F2026-09-20\u002Fdoc-inisnssw6528969.shtml","西南林业大学李传孟、杨洁副教授、张晓宇在《智慧农业（中英文）》2026,8(4):85-99发表Flor-YOLO模型，针对香石竹鲜切花开放度人工分级主观性强、效率低问题，以YOLO11n为基线进行骨干网络、下采样方式、检测头结构针对性改进。Flor-YOLO在自建香石竹数据集上mAP@50达到96.10%，较基准模型提升3.25个百分点；模型参数量与计算量分别为1.26M和1.1GFLOPs，同比降低51.2%和82.5%。","智慧农业(中英文)2026,8(4):85-99","2026-09-20T12:26:00Z",25,83,{"impact":19,"substance":274,"depth":19,"authority":108,"freshness":13,"relevant":22,"comment":393},"核心期刊论文，提出轻量化Flor-YOLO模型实现香石竹鲜切花自动分级，数据翔实、方法有创新，对花卉产业智能化有参考价值。",[395,396],{"name":388,"url":386},{"name":397,"url":398},"《智慧农业（中英文）》2026年第4期","https:\u002F\u002Fwww.toutiao.com\u002Farticle\u002F7687597315586818611",[27,28,400,31,401],"花卉产业","鲜切花分级",{"code":40,"slug":228,"name":229,"path":230,"parent_code":231},[],{"status":104,"locked":14,"version":35,"method":36,"evidence":405},{"candidates":406,"agricultural_evidence":55},[407,411,416],{"code":40,"score":20,"supported":55,"title":408,"abstract":409,"tags":410},[292],[292],[28,31],{"code":51,"score":412,"supported":55,"title":413,"abstract":414,"tags":415},4,[],[27],[27],{"code":417,"score":22,"supported":14,"title":418,"abstract":419,"tags":420},"B4",[],[],[400],[422,423],"西南林业大学 香石竹 鲜切花分级","Flor-YOLO 香石竹 检测","西南林业大学香石竹鲜切花分级-3046",{"doi":9,"openalex_id":9,"authors":426,"venue":9,"cited_by_count":69,"oa_url":9,"card":427,"direction":74,"ingested_from":208},[],{"tldr":428,"method":429,"finding":430,"direction":74,"opportunity":431},"提出轻量化Flor-YOLO模型，实现香石竹鲜切花开放度自动分级检测。","以YOLO11n为基线，改进骨干网络、下采样方式与检测头，自建香石竹数据集。","mAP@50达96.10%，较基准提升3.25个百分点，参数量与计算量分别降低51.2%和82.5%","可迁移至其他花卉或果蔬的轻量化分级，并探索边缘设备实时部署与多任务联合检测。","2026-09-21T00:04:39.154232Z"]