[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"item-470":3,"related-470":56},{"id":4,"title":5,"url":6,"summary":7,"summary_zh":8,"content":9,"source_name":10,"source_url":8,"published_at":11,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":16,"sources":23,"tags":25,"primary_topic":30,"secondary_topics":36,"classification":37,"search_phrases":50,"slug":53,"view_count":54,"doi":8,"paper":8,"created_at":55},470,"A TinyMLOps Pipeline for Coarse-Grained Plant Disease Classification in Precision Agriculture","https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029","本文提出TinyMLOps管道,涵盖模型设计、优化、量化及跨多种边缘设备部署,在受控实验室条件下评估。基于PlantVillage基准推导的数据集,将39个细粒度类别聚合成3个超类(健康叶、不健康叶、无叶),系统执行植物健康状态分类和背景过滤功能,而不是诊断特定疾病。训练基于MobileNet的卷积神经网络,联合优化分类精度和计算效率;选择5个模型(Pareto前沿4个+最大HPO模型)转换为LiteRT和ONNX,在Raspberry Pi Zero 2 W和STM32H743ZI微控制器上以float32和后训练int8精度评估。在float32下,LiteRT比ONNX Runtime快1.87-2.65倍;int8 LiteRT模型比float32小2.83-3.67倍,快21.3-31.2%,F1-score损失仅0.0010-0.0068;在微控制器上,两个最小模型int8变体比float32快4.3倍,MCU闪存占用小3.85倍。",null,"Logical Operator Operator\n\nSearch Text\n\nSearch Type\n\n_add\\_circle\\_outline_\n\n_remove\\_circle\\_outline_\n\n[![Image 1: sensors-logo](https:\u002F\u002Fpub.mdpi-res.com\u002Fimg\u002Fjournals\u002Fsensors-logo.png?e6c5d1b63c26048a)](https:\u002F\u002Fwww.mdpi.com\u002Fjournal\u002Fsensors)\n\n## Article Menu\n\nFont Type:\n\n_Arial_ _Georgia_ _Verdana_\n\nFont Size:\n\nAa Aa Aa\n\nLine Spacing:\n\n__ __ __\n\nColumn Width:\n\n__ __ __\n\nBackground:\n\nOpen Access Article\n\nby \nHossein Aqasizade\n\n 1,2[](mailto:haqasizade@fbk.eu)[![Image 2: ORCID](https:\u002F\u002Fpub.mdpi-res.com\u002Fimg\u002Fdesign\u002Forcid.png?0465bc3812adeb52?1786452582)](https:\u002F\u002Forcid.org\u002F0009-0008-4193-0470), \nMattia Antonini\n\n 1[](mailto:m.antonini@fbk.eu)[![Image 3: ORCID](https:\u002F\u002Fpub.mdpi-res.com\u002Fimg\u002Fdesign\u002Forcid.png?0465bc3812adeb52?1786452582)](https:\u002F\u002Forcid.org\u002F0000-0003-1685-0264), \nMassimo Vecchio\n\n 1,*[](mailto:mvecchio@fbk.eu)[![Image 4: ORCID](https:\u002F\u002Fpub.mdpi-res.com\u002Fimg\u002Fdesign\u002Forcid.png?0465bc3812adeb52?1786452582)](https:\u002F\u002Forcid.org\u002F0000-0003-4426-8220) and \nFabio Antonelli\n\n 1[](mailto:fantonelli@fbk.eu)[![Image 5: ORCID](https:\u002F\u002Fpub.mdpi-res.com\u002Fimg\u002Fdesign\u002Forcid.png?0465bc3812adeb52?1786452582)](https:\u002F\u002Forcid.org\u002F0000-0001-5221-7752)\n\n1\n\nFondazione Bruno Kessler, 38123 Trento, Italy\n\n2\n\nDepartment of Information Engineering and Computer Science, University of Trento, 38123 Trento, Italy\n\n*\n\nAuthor to whom correspondence should be addressed.\n\nSubmission received: 23 May 2026 \u002F Revised: 31 July 2026 \u002F Accepted: 4 August 2026 \u002F Published: 7 August 2026\n\n## Abstract\n\nIdentifying plant health conditions is an emerging precision-agriculture and food-security challenge, intensified by deploying deep-learning models on memory- and power-constrained edge devices. We present a TinyMLOps pipeline spanning model design, optimization, quantization, and deployment across diverse edge devices, evaluated under controlled laboratory conditions. Using a dataset derived from the PlantVillage benchmark, 39 fine-grained classes are aggregated into three superclasses: healthy leaf, unhealthy leaf, and no leaves. The resulting system therefore performs plant health-status classification and background filtering rather than diagnosing specific diseases. We train a MobileNet-based convolutional neural network jointly optimized for classification accuracy and computational efficiency, adopting state-of-the-art hyperparameter optimization (HPO) tools. Five models are selected, four from the Pareto Front and one as the biggest evaluated model during HPO, converted to LiteRT and ONNX, and evaluated at\n\nfloat32\nand post-training\n\nint8\nprecision on a Raspberry Pi Zero 2 W and an STM32H743ZI microcontroller. At\n\nfloat32\n, LiteRT is 1.87–2.65× faster than ONNX Runtime on the Raspberry Pi across all five models. Relative to their\n\nfloat32\nLiteRT counterparts, the\n\nint8\nLiteRT models are 2.83–3.67× smaller on disk and 21.3–31.2% faster on the same board, at a cost in F1-score of between 0.0010 and 0.0068. On the microcontroller, only the two smallest models deploy at both precisions; for these, the fully quantized\n\nint8\nvariants are\n\n$4.3 \\times$\nfaster and 3.85× smaller in MCU flash footprint than the\n\nfloat32\ncounterparts. The mid-range model fits the 2 MB flash and 1 MB RAM budget only when quantized, while the two largest models exceed it in every configuration tested.\n\n## 1. Introduction\n\nAs the need for food security grows, precision farming stands as a powerful ally, leveraging the most cutting-edge technologies to enhance crop production. Every year, significant portions of farm production are wasted due to recurring pests, devastating diseases, harsh weather patterns, and subtle soil erosion, all of which contribute to these losses. The precision farming paradigm, enabled by modern technology to make farming more efficient and productive, provides a powerful answer to these growing threats. To protect plants and secure the farming economy, the ability to make early, accurate plant disease identification remains fundamental, as it enables quick action before a widespread epidemic. Traditionally, this function was provided by biological methods, as in the case of “sentinel” plants [[1](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B1-sensors-26-05029)], which exhibit symptoms indicative of particular pathogens, serving as a sort of natural early warning system. Although theoretically sound, traditional methods are bogged down by practical constraints. They are quite time- and labor-intensive to plant, grow, and continuously monitor. Further, human visual inspection is prone to error, particularly in misinterpreting or completely overlooking disease symptoms [[2](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B2-sensors-26-05029)]. As a countermeasure to the limitations of manual, biology-focused monitoring, the farming industry continues to embrace technological innovations. Digital farming leverages solutions such as computer vision, Internet of Things (IoT) sensor systems, and AI-driven predictive analytics to provide faster, scalable, and more reliable real-time monitoring of crop health [[3](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B3-sensors-26-05029)]. Among the array of methods, the combination of computer vision and deep learning has proven one of the most effective for mechanizing plant disease classification from images [[4](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B4-sensors-26-05029)].\n\nThe noteworthy aspect in this case is the direct execution on edge devices, from microcontrollers (MCUs) and single-board computers (SBCs) to real-time, in-situ processing of a continuous stream. Such embedded settings are themselves defined by tight computational constraints that profoundly impact model design, from architectural size (megabytes vs. kilobytes) to data types (\n\nfloat32\nvs.\n\nint8\n), among others. In real agricultural deployments, connectivity is often intermittent or entirely unavailable, and power budgets are limited. For this reason, disease detection is better performed locally, directly on embedded devices, enabling real-time responses while minimizing communication, energy consumption, and operational cost. Systematically managing these restrictions is necessary, and the TinyMLOps approach provides a systematic methodology for model design, optimization, and execution on devices with limited capabilities [[5](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B5-sensors-26-05029)].\n\nThis work describes the development of a computer vision system based on TinyMLOps principles, designed to run an optimized deep neural network model on embedded platforms for plant disease monitoring. The system uses the state-of-the-art MobileNetV1 backbone architecture [[6](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B6-sensors-26-05029)] as a pre-trained feature extractor, while a custom final classification layer is trained from scratch. Optuna [[7](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B7-sensors-26-05029)] is utilized for hyperparameter tuning, balancing detection accuracy and model complexity.\n\nThe experiments were conducted on the PlantVillage dataset, a widely used benchmark in the plant disease detection literature, consisting of high-quality leaf images acquired under controlled laboratory conditions. While this dataset supports rigorous and reproducible evaluation of the vision-sensing pipeline, it does not reflect the full complexity of outdoor deployments, including variable lighting, occlusions, and background clutter. The quantitative results reported here should therefore be interpreted as benchmark performance. To assess operability beyond PlantVillage, the system is additionally evaluated qualitatively on previously unseen, locally collected leaves ([Section 4.2.3](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec4dot2dot3-sensors-26-05029)). This work contributes the systematic integration of model optimization, deployment, and hardware evaluation into a reproducible TinyMLOps lifecycle for heterogeneous edge devices in precision agriculture. The ultimate operational objective is real-time, on-device plant-health screening under field conditions. The envisioned inference pipeline for field deployment is illustrated in [Figure 1](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#fig_body_display_sensors-26-05029-f001).\n\n**Figure 1.** Envisioned on-device inference pipeline for real-time plant-health screening under field conditions. A captured image is preprocessed and passed to the quantized model, which classifies the input as a healthy leaf, an unhealthy leaf, or no leaf.\n\nThe main contributions of this article are summarized as follows:\n\n1.\nDesigned and implemented an end-to-end TinyMLOps pipeline for edge deployment, integrating state-of-the-art lightweight CNN architectures (e.g., MobileNet) and covering the full workflow from model development to on-device inference;\n\n2.\nComprehensive hardware-level benchmarking and deployment validation on heterogeneous edge devices, namely a Raspberry Pi Zero 2 W and an STM32H743ZI microcontroller, providing detailed analysis of inference latency, memory footprint, and accuracy trade-offs, and establishing the feasibility of on-device inference.\n\n3.\nOptimization and adaptation of plant disease classification models for multiple resource-constrained devices.\n\nThe remaining parts of the paper are organized as follows: [Section 2](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec2-sensors-26-05029) provides an overview of research in Edge AI and TinyMLOps, reviewing prior frameworks, deployments, and optimization approaches. [Section 3](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec3-sensors-26-05029) details the proposed workflow, including dataset preparation, model design\u002Ftraining, optimization, and edge-device deployment and configuration. [Section 4](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec4-sensors-26-05029) reports the experimental results on both target platforms, together with a discussion of the observed trade-offs, limitations, and a roadmap towards field deployment. Finally, [Section 5](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec5-sensors-26-05029) concludes the paper by summarizing the main findings and outlining directions for future work.\n\n## 2. Related Work\n\nResearch related to this study spans multiple intersecting areas, including TinyMLOps frameworks, edge AI architectures, model and optimizations for TinyML, and plant disease detection on edge devices. This section reviews the most relevant contributions across these dimensions and positions the proposed work within the existing literature.\n\n### 2.1. Vision-Based Plant Disease Detection in Precision Agriculture\n\nWadhwa et al. [[8](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B8-sensors-26-05029)] propose a generalized hybrid CNN ensemble for multi-crop disease classification that uses F1-score driven Dirichlet weighting and CatBoost stacking, with SHAP-based explainability, while Sakthipriya et al. [[9](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B9-sensors-26-05029)] propose a precision agriculture approach for rice nutrient management that uses genetic algorithms to automatically design and optimize CNN architectures. Their GENCNN framework focuses on model automation and accuracy, aiming to improve nitrogen management and reduce fertilizer costs, but does not address deployment, edge constraints, or operational lifecycle aspects. Sivakumar et al. [[10](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B10-sensors-26-05029)] propose a computer vision–based crop and weed segmentation framework using federated ensemble learning across heterogeneous multi-crop field datasets. Their approach enables scalable, privacy-preserving model training and incremental deployment without centralized data sharing. Khan et al. [[11](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B11-sensors-26-05029)] propose a MobileNetV3-Small-based model for plant disease classification on edge devices. Their approach focuses on achieving high accuracy on the PlantVillage dataset and applies post-training quantization to reduce model size. Several additional studies provide relevant accuracy–complexity references on PlantVillage, including classifiers based on AlexNet and GoogLeNet [[12](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B12-sensors-26-05029)], InceptionV3 [[13](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B13-sensors-26-05029)], a compact nine-layer CNN [[14](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B14-sensors-26-05029)], and a modified InceptionV3 [[15](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B15-sensors-26-05029)]. Their reported predictive metrics and parameter counts are interpreted in [Section 4.4](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#sec4dot4-sensors-26-05029), with explicit consideration of differences in class granularity and experimental protocol. Arratia et al. [[16](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B16-sensors-26-05029)] propose the BODOQUE (Bimodal Observational Device for Optimizing Quantification of Ephemeral streams), a systematic framework for deploying TinyML in real-time environmental monitoring that combines low-power sensing with event-triggered, high-performance processing to minimize energy consumption. Other agricultural applications emphasize model design rather than operational deployment. In [[17](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B17-sensors-26-05029)], the authors introduced a general inference pipeline for plant disease detection running on a small single-board computer (i.e., a Raspberry Pi Zero 2 W). Building on that foundation, this paper focuses on the specific optimizations required for efficient on-device execution at the network edge. Although these studies demonstrate the effectiveness of deep learning for agricultural disease monitoring, many prioritize classification accuracy over deployment feasibility, runtime efficiency, and operational lifecycle management on embedded sensing platforms.\n\n### 2.2. Lightweight Deep Learning Models for Edge Agricultural Monitoring\n\nTo go beyond hardware limits, scientists have attempted architectural and algorithmic advancements. Larian et al. [[18](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B18-sensors-26-05029)] introduced the InTec framework and experimented with a three-layer architecture. They found that dividing the ML pipeline across the thing, edge, and cloud layers might significantly reduce latency and network traffic. In a more comprehensive study, Jenhani et al. [[19](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B19-sensors-26-05029)] conducted an experimental investigation of split learning. On a customized testbed composed of ESP32 boards, they experimented with the efficiency of a partitioned MobileNetV2 model across different low-power wireless protocols. They identified optimal split points to balance on-device computation and communication overhead. Rather than disseminating the model, Hayajneh et al. [[20](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B20-sensors-26-05029)] examined the potential to enhance the model’s adaptability. Their study on on-device transfer learning demonstrated its feasibility for enabling models to acquire knowledge from novel data directly at the edge. Cerioli et al. [[21](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B21-sensors-26-05029)] expressly stated the advantage of strongly specialized optimization in their research on the detection of microplastics. They built a tailored compiler to generate native C code for their model without relying on standard inference engines. From their experimental data, they found that their technique achieved significant increases in latency, memory use, and energy efficiency compared with typical techniques. Similarly, Lê et al. [[22](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B22-sensors-26-05029)] are interested in the run-time dynamics of TinyML and introduce a new TinyMLOps approach to real-time frame-based event classification on ultra-low-power microcontrollers. They presented a comparison using available frameworks, such as LiteRT and NNoM, to support their strategy. They even use TinyML to secure consumer IoT devices.\n\n### 2.3. TinyMLOps and Edge AI Frameworks for Resource-Constrained Deployment\n\nThe Edge AI paradigm is a revolutionary shift that is driving new approaches to deploying applications on resource-limited devices, especially in verticals such as precision agriculture. The peculiar restrictions on these systems have led to the definition of Tiny Machine Learning Operations (TinyMLOps) [[5](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B5-sensors-26-05029)]. Antonini et al. research [[5](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B5-sensors-26-05029)] presents a comprehensive, end-to-end industrial IoT application for operationalizing anomaly detection in submersible water pumps. Additionally, this paper utilizes readily available, inexpensive hardware and presents a rigorous empirical study to support its findings. Moreover, they provide a formal, detailed description of the conceptual TinyMLOps framework, which applies classical machine learning orchestration principles to the unique domain of non-Linux-based, far-edge devices. However, Leroux et al. [[23](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B23-sensors-26-05029)] say wide deployment is impeded by operational obstacles that transcend computation and reach to model observation, management, security, and integrity verification on distributed device populations. To address these operational concerns, various MLOps frameworks have been developed.\n\nFor an open-source alternative, Peltonen et al. suggested LinkEdge [[24](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B24-sensors-26-05029)], an end-to-end toolset for connecting edge devices and supporting MLOps. They tested the system by creating a model to predict air-pressure failures in trucks and concluded that their open-source alternative performed as well as commercial tools while providing freedom from vendor lock-in. Extending on-device execution boundaries, Min et al. [[25](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B25-sensors-26-05029)] created a multi-tenant sensory edge device runtime called SensiX++. They discovered that a resource-aware model server and an adaptive scheduler could be adopted to securely and efficiently run multiple isolated models on a single device. To overcome the practical hurdles across the entire workflow, Huang et al. [[26](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B26-sensors-26-05029)] proposed the RIOT-ML toolkit. They propose a “batteries-included” package that includes a low-power OS, a model transpiler, and a secure over-the-air (OTA) update system and demonstrate its usefulness for benchmarking model performance on diverse IoT hardware. In other contributions, Karras et al. [[27](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B27-sensors-26-05029)] contend that effective data management is a condition for successful edge AI. Towards that end, they designed and delivered a suite of TinyML-specific data-cleaning, clustering, and compression algorithms.\n\nIn other work, Katib et al. [[28](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B28-sensors-26-05029)] designed a Deep Learning technique to secure IoT consumer devices with TinyML Driven Real-time Anomaly Detection for Predictive Maintenance (DLTML-RTADPM), to achieve real-time anomaly detection and predictive maintenance. Their approach comprises Z-score normalization as the data preprocessing method, the Fennec Fox Optimization Algorithm to select features, a bidirectional long short-term memory (BiLSTM) network for anomaly detection, and the Jaya optimization algorithm to optimize hyperparameters. Flores et al. [[29](https:\u002F\u002Fwww.mdpi.com\u002F1424-8220\u002F26\u002F16\u002F5029#B29-sensors-26-05029)] suggest a system of tiny machine learning tasks under the Internet of Intelligent Vehicles. Two ESP32 units are used in their solution, with one serving as a radio station and the other as an onboard diagnostic scanner. These updated models are then sent over a wireless channel. This system shows a marked improvement in model efficacy pos","MDPI Sensors 2026-08-07","2026-08-07T00:00:00Z","论文",10,false,67,{"impact":17,"substance":18,"depth":19,"authority":17,"freshness":20,"relevant":21,"comment":22},12,20,18,5,1,"论文提出TinyMLOps流水线，在边缘设备上实现植物健康状态分类，方法新颖且实验详实，对智慧农业边缘计算有参考价值。",[24],{"name":10,"url":6},[26,27,28,29],"精准农业","TinyML","边缘AI","植物病害",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},"I1","smart-agriculture","智慧农业总论","\u002Ftopic\u002Fdigital-agriculture\u002Fsmart-agriculture","I",[],{"status":38,"locked":14,"version":39,"method":40,"evidence":41},"classified","2026-10-22.2","rules",{"candidates":42,"agricultural_evidence":45},[43],{"code":31,"score":44,"supported":45,"title":46,"abstract":48,"tags":49},9,true,[47],"precision agriculture",[],[26],[51,52],"植物病害 精准农业 边缘AI TinyML","植物病害 精准农业","植物病害精准农业边缘AITinyML-470",3,"2026-08-12T23:57:24.355943Z",{"total":57,"page":21,"page_size":57,"items":58},6,[59,111,163,206,252,308],{"id":60,"title":61,"url":62,"summary":63,"summary_zh":8,"content":8,"source_name":64,"source_url":8,"published_at":65,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":66,"score_detail":67,"sources":71,"tags":73,"primary_topic":77,"secondary_topics":78,"classification":79,"search_phrases":96,"slug":99,"view_count":21,"doi":8,"paper":100,"created_at":110},4647,"[预印本]植物病害数据集的数据中心综述:分类、关键分析、环境变异性与精准农业启示(A Data-Centric Review of Plant Disease Datasets: Taxonomy, Critical Analysis, Environmental Variability, and Implications for Precision Agriculture)","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.07087","arXiv综述(2610.07087,2026年10月5日提交)从数据采集、可访问性、植物多样性、病害构成、类别结构与不平衡严重度建立植物病害数据集分类法,对比实验室与田间数据集差异,分析类内、跨作物、跨数据集多层次不平衡及环境变异性对模型性能的影响,提出融合温湿度、叶面湿润度等多模态环境参数的方向。[预印本未经同行评议]。","arXiv 2026-10-05 [预印本]","2026-10-05T00:00:00Z",71,{"impact":68,"substance":69,"depth":19,"authority":44,"freshness":44,"relevant":21,"comment":70},14,21,"arXiv预印本系统梳理植物病害数据集分类与不平衡问题，对农业AI数据建设有参考价值，但未经同行评议且偏学术，适合主题聚合而非头条精选。",[72],{"name":64,"url":62},[74,75,26,29,76],"智慧农业","农业人工智能","农业数据集",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":80},{"candidates":81,"agricultural_evidence":45},[82,86,91],{"code":31,"score":19,"supported":45,"title":83,"abstract":84,"tags":85},[26,47],[],[74,26],{"code":87,"score":21,"supported":14,"title":88,"abstract":89,"tags":90},"I2",[],[],[75],{"code":92,"score":21,"supported":14,"title":93,"abstract":94,"tags":95},"I6",[],[],[76],[97,98],"植物病害 环境变异性 精准农业","农业人工智能 农业数据集 智慧农业 植物病害","植物病害环境变异性精准农业-4647",{"doi":8,"openalex_id":8,"authors":101,"venue":8,"cited_by_count":102,"oa_url":8,"card":103,"direction":107,"ingested_from":109},[],0,{"tldr":104,"method":105,"finding":106,"direction":107,"opportunity":108},"综述植物病害数据集分类与不平衡问题，提出融合环境参数的方向。","综述数据集分类法，对比实验室与田间数据，分析多层次不平衡。","环境变异性和多层次类别不平衡显著影响病害识别模型性能。","农业人工智能与决策模型","构建融合温湿度、叶面湿润度等环境参数的多模态病害数据集与鲁棒模型。","agent","2026-10-11T00:09:47.731337Z",{"id":112,"title":113,"url":114,"summary":115,"summary_zh":8,"content":8,"source_name":116,"source_url":114,"published_at":117,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":118,"score_detail":119,"sources":122,"tags":124,"primary_topic":127,"secondary_topics":128,"classification":129,"search_phrases":141,"slug":144,"view_count":102,"doi":145,"paper":146,"created_at":162},5539,"Robust and transparent crop-row detection from RGB field imagery for vision-based agricultural guidance systems","https:\u002F\u002Fdoi.org\u002F10.1007\u002Fs11119-026-10473-z","Robust and transparent crop-row detection from RGB field imagery for vision-based agricultural guidance systems。Precision Agriculture","Precision Agriculture","2026-10-22T00:00:00Z",79,{"impact":120,"substance":69,"depth":19,"authority":68,"freshness":13,"relevant":21,"comment":121},16,"核心期刊新论文，提出鲁棒且可解释的RGB作物行检测方法，对视觉导航农机具参考价值明确，但属细分技术进展，未达重大突破层级。",[123],{"name":116,"url":114},[74,75,26,125,126],"机器视觉","作物行检测",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":130},{"candidates":131,"agricultural_evidence":45},[132,136],{"code":31,"score":20,"supported":45,"title":133,"abstract":134,"tags":135},[],[47],[74,26],{"code":87,"score":137,"supported":14,"title":138,"abstract":139,"tags":140},2,[],[],[75,125],[142,143],"RGB 作物行检测","农业人工智能 作物行检测 智慧农业 机器视觉","RGB作物行检测-5539","10.1007\u002Fs11119-026-10473-z",{"doi":145,"openalex_id":147,"authors":148,"venue":116,"cited_by_count":102,"oa_url":8,"card":8,"direction":8,"ingested_from":161},"W7223117828",[149,152,155,158],{"name":150,"orcid":151},"Esther Gascó","https:\u002F\u002Forcid.org\u002F0000-0003-1798-4483",{"name":153,"orcid":154},"Clara I. López-González","https:\u002F\u002Forcid.org\u002F0000-0002-2755-1692",{"name":156,"orcid":157},"Eva Besada-Portas","https:\u002F\u002Forcid.org\u002F0000-0001-6129-4653",{"name":159,"orcid":160},"Gonzalo Pájares","https:\u002F\u002Forcid.org\u002F0000-0003-0915-6282","openalex","2026-10-22T11:20:09.678587Z",{"id":164,"title":165,"url":166,"summary":167,"summary_zh":8,"content":8,"source_name":168,"source_url":166,"published_at":169,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":170,"score_detail":171,"sources":174,"tags":176,"primary_topic":179,"secondary_topics":180,"classification":181,"search_phrases":194,"slug":197,"view_count":102,"doi":198,"paper":199,"created_at":205},5380,"Automation, Robotics, and Precision Agriculture Transformation Through Smart Harvest Systems","https:\u002F\u002Fdoi.org\u002F10.13140\u002Frg.2.2.17987.98089","Automation, Robotics, and Precision Agriculture Transformation Through Smart Harvest Systems。OpenAlex","OpenAlex","2026-10-19T00:00:00Z",56,{"impact":19,"substance":172,"depth":13,"authority":17,"freshness":172,"relevant":21,"comment":173},8,"主题契合智慧农业，但摘要仅为标题重复、缺乏方法与数据细节，信息增量有限，暂不建议进入每日精选。",[175],{"name":168,"url":166},[74,177,26,178],"农业机器人","智能采收",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":182},{"candidates":183,"agricultural_evidence":45},[184,189],{"code":31,"score":185,"supported":45,"title":186,"abstract":187,"tags":188},13,[47],[47],[74,26],{"code":190,"score":21,"supported":14,"title":191,"abstract":192,"tags":193},"F6",[],[],[177],[195,196],"农业机器人 智慧农业 智能采收 精准农业","农业机器人 智慧农业","农业机器人智慧农业智能采收精准农业-5380","10.13140\u002Frg.2.2.17987.98089",{"doi":198,"openalex_id":200,"authors":201,"venue":8,"cited_by_count":102,"oa_url":166,"card":8,"direction":204,"ingested_from":161},"W7222971426",[202],{"name":203,"orcid":8},"Joydeep Dey","智慧农业 \u002F 农业物联网","2026-10-20T23:30:08.600540Z",{"id":207,"title":208,"url":209,"summary":210,"summary_zh":211,"content":8,"source_name":212,"source_url":209,"published_at":169,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":118,"score_detail":213,"sources":216,"tags":218,"primary_topic":221,"secondary_topics":222,"classification":223,"search_phrases":235,"slug":238,"view_count":102,"doi":239,"paper":240,"created_at":251},5316,"PLANT-GPK v2.7: Spatial Heterogeneity and Predictive Digital Twin for Plant Phenotyping","https:\u002F\u002Fdoi.org\u002F10.5281\u002Fzenodo.23460273","Traditional image-based plant phenotyping assumes spatial homogeneity, estimating a single set of physiological parameters for the entire plant. This paper introduces a novel extension to the PLANT-GPK framework: Spatial Physiological Heatmaps and Predictive Digital Twins. By dividing the plant image into a 3x3 grid and solving the inverse ODE problem for each spatial patch, we reveal the hidden spatial heterogeneity of resource allocation. Furthermore, the recovered parameters are extrapolated to predict the exact day of harvest using a continuous smooth-physics model, providing a powerful prescriptive tool for precision agriculture.","传统的基于图像的植物表型分析假设空间同质性，即对整个植株估计单一组生理参数。本文提出了对PLANT-GPK框架的一种新颖扩展：空间生理热图与预测性数字孪生。通过将植物图像划分为3×3网格并针对每个空间区块求解逆常微分方程（ODE）问题，我们揭示了资源分配中隐藏的空间异质性。此外，利用连续光滑物理模型对所恢复的参数进行外推，以预测确切的收获日期，为精准农业提供了强有力的规范性工具。","Zenodo (CERN European Organization for Nuclear Research)",{"impact":19,"substance":214,"depth":19,"authority":17,"freshness":44,"relevant":21,"comment":215},22,"该论文提出植物表型空间异质性建模与预测性数字孪生方法，方法新颖、结论有实用价值，但属细分领域预印本成果，未达重大突破层级。",[217],{"name":212,"url":209},[74,75,219,26,220],"数字孪生","植物表型",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":224},{"candidates":225,"agricultural_evidence":45},[226,231],{"code":31,"score":18,"supported":45,"title":227,"abstract":229,"tags":230},[228],"digital twin",[26,219,47],[219,74,26],{"code":87,"score":21,"supported":14,"title":232,"abstract":233,"tags":234},[],[],[75],[236,237],"PLANT-GPK 植物表型","空间生理热图 数字孪生","PLANT-GPK植物表型-5316","10.5281\u002Fzenodo.23460273",{"doi":239,"openalex_id":241,"authors":242,"venue":212,"cited_by_count":102,"oa_url":209,"card":245,"direction":249,"ingested_from":161},"W7222885652",[243],{"name":244,"orcid":8},"MOHAMMED TAHAR SAIM",{"tldr":246,"method":247,"finding":248,"direction":249,"opportunity":250},"将植物图像分3x3网格反演ODE参数，构建空间生理热图与预测数字孪生。","基于PLANT-GPK框架，分块求解逆ODE问题，并用连续平滑物理模型外推。","揭示了植物资源分配的空间异质性，可预测精确收获日。","农业遥感与作物表型","可探索更细粒度网格与多模态表型融合，提升数字孪生预测精度与田间适用性。","2026-10-19T23:30:19.414490Z",{"id":253,"title":254,"url":255,"summary":256,"summary_zh":257,"content":8,"source_name":258,"source_url":255,"published_at":259,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":260,"score_detail":261,"sources":263,"tags":265,"primary_topic":270,"secondary_topics":271,"classification":272,"search_phrases":284,"slug":287,"view_count":21,"doi":288,"paper":289,"created_at":307},5103,"Assessment of the efficiency of resource potential utilization in the agricultural sector of the economy under investment and innovation development","https:\u002F\u002Fdoi.org\u002F10.32782\u002Fnvlvet-e10701","Ukraine’s agricultural sector is undergoing transformation under the influence of wartime challenges, economic restructuring, European integration, digitalization, and the growing need for sustainable resource management. Improving the efficiency of resource potential utilization has become a prerequisite for strengthening the competitiveness and sustainability of agricultural enterprises. The purpose of this study is to provide a comprehensive assessment of the efficiency of resource potential utilization in the agricultural sector of the economy under investment and innovation development and to substantiate organizational and economic measures for its improvement. The research is based on a systems approach using economic-statistical, comparative, analytical, economic-mathematical, graphical, forecasting, and composite assessment methods. The empirical analysis covers agricultural enterprises in the Poltava region during 2015–2025. Resource potential utilization efficiency was comprehensively assessed by integrating economic, investment, innovation, environmental, and managerial indicators. The results show that technological modernization, digital transformation, investment activity, and strategic management are the principal drivers of efficient resource utilization and sustainable enterprise development. An integrated organizational and economic model for resource potential development is proposed, combining precision agriculture, digital monitoring, resource-saving technologies, and human capital development. Forecasting confirms the positive impact of the proposed measures on labour productivity, capital productivity, profitability, investment activity, digitalization, and the integrated indicator of resource utilization efficiency. The scientific novelty lies in improving a comprehensive methodological approach to assessing the efficiency of resource potential utilization in the agricultural sector of the economy through integrating economic, investment, innovation, environmental, and managerial components. The practical significance lies in applying the proposed approach to strategic planning for agricultural enterprise development, investment decision-making, and improving resource management efficiency.","乌克兰农业部门正处在战时挑战、经济结构调整、欧洲一体化、数字化以及可持续资源管理需求日益增长等多重因素影响下的转型进程之中。提高资源潜力利用效率已成为增强农业企业竞争力和可持续性的先决条件。本研究旨在对投资与创新发展背景下经济中农业部门的资源潜力利用效率进行综合评估，并为其改进提供组织与经济措施的依据。研究基于系统方法，运用了经济统计、比较、分析、经济数学、图形、预测及综合评估等方法。实证分析涵盖2015—2025年波尔塔瓦地区的农业企业。通过整合经济、投资、创新、环境和管理指标，对资源潜力利用效率进行了综合评估。结果表明，技术现代化、数字化转型、投资活动和战略管理是高效资源利用与企业可持续发展的主要驱动力。研究提出了一个资源潜力发展的综合组织与经济模型，将精准农业、数字监测、资源节约技术和人力资本开发相结合。预测结果证实了所提措施对劳动生产率、资本生产率、盈利能力、投资活动、数字化以及资源利用效率综合指标的积极影响。本研究的科学新颖性在于，通过整合经济、投资、创新、环境和管理等组成部分，改进了评估经济中农业部门资源潜力利用效率的综合方法论。其实践意义在于将所提方法应用于农业企业发展的战略规划、投资决策以及提高资源管理效率。","Scientific Messenger of LNU of Veterinary Medicine and Biotechnologies","2026-10-15T00:00:00Z",63,{"impact":172,"substance":19,"depth":120,"authority":17,"freshness":44,"relevant":21,"comment":262},"区域实证研究，方法综合但影响限于乌克兰波尔塔瓦地区，对国内三农信息化参考价值有限。",[264],{"name":258,"url":255},[266,26,267,268,269],"数字农业","农业投资","乌克兰农业","农业资源利用",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":273},{"candidates":274,"agricultural_evidence":45},[275,279],{"code":31,"score":172,"supported":45,"title":276,"abstract":277,"tags":278},[],[26,47],[266,26],{"code":280,"score":21,"supported":14,"title":281,"abstract":282,"tags":283},"G3",[],[],[267],[285,286],"波尔塔瓦 农业企业 资源潜力","乌克兰 数字化 投资","波尔塔瓦农业企业资源潜力-5103","10.32782\u002Fnvlvet-e10701",{"doi":288,"openalex_id":290,"authors":291,"venue":258,"cited_by_count":102,"oa_url":255,"card":301,"direction":305,"ingested_from":161},"W7221969015",[292,295,298],{"name":293,"orcid":294},"Olena Bukhalo","https:\u002F\u002Forcid.org\u002F0000-0001-7776-6715",{"name":296,"orcid":297},"Андрій Петрович Дорошенко","https:\u002F\u002Forcid.org\u002F0000-0002-6314-1586",{"name":299,"orcid":300},"Yuri Nesterenko","https:\u002F\u002Forcid.org\u002F0009-0000-5831-4465",{"tldr":302,"method":303,"finding":304,"direction":305,"opportunity":306},"评估乌克兰波尔塔瓦地区农业企业资源潜力利用效率并提出改进措施。","系统方法结合经济统计、比较、预测与复合评估，分析2015-2025年农业企业数据","技术现代化、数字化、投资与战略管理是资源高效利用和可持续发展的主要驱动力。","数字乡村与农业信息化","可延伸至战后或转型期农业数字化投资对资源效率的因果机制与动态监测研究。","2026-10-16T23:30:33.662938Z",{"id":309,"title":310,"url":311,"summary":312,"summary_zh":8,"content":8,"source_name":313,"source_url":311,"published_at":314,"category":12,"cover_url":8,"hotness":13,"is_selected":14,"score":15,"score_detail":315,"sources":318,"tags":320,"primary_topic":322,"secondary_topics":323,"classification":324,"search_phrases":340,"slug":343,"view_count":137,"doi":344,"paper":345,"created_at":359},5082,"Entomopathogenic nematode recent applications in precision agriculture: Methods, challenges, and future perspectives","https:\u002F\u002Fdoi.org\u002F10.1016\u002Fj.jip.2026.108765","Entomopathogenic nematode recent applications in precision agriculture: Methods, challenges, and future perspectives。Journal of Invertebrate Pathology","Journal of Invertebrate Pathology","2026-10-14T00:00:00Z",{"impact":17,"substance":120,"depth":316,"authority":68,"freshness":172,"relevant":21,"comment":317},17,"核心期刊综述，聚焦昆虫病原线虫在精准农业中的应用方法与挑战，专业性强但属细分领域，产业影响有限。",[319],{"name":313,"url":311},[74,75,26,321],"生物防治",{"code":31,"slug":32,"name":33,"path":34,"parent_code":35},[],{"status":38,"locked":14,"version":39,"method":40,"evidence":325},{"candidates":326,"agricultural_evidence":45},[327,331,336],{"code":31,"score":185,"supported":45,"title":328,"abstract":329,"tags":330},[47],[47],[74,26],{"code":332,"score":21,"supported":14,"title":333,"abstract":334,"tags":335},"F4",[],[],[321],{"code":87,"score":21,"supported":14,"title":337,"abstract":338,"tags":339},[],[],[75],[341,342],"昆虫病原线虫 精准农业","Journal of Invertebrate Pathology 线虫","昆虫病原线虫精准农业-5082","10.1016\u002Fj.jip.2026.108765",{"doi":344,"openalex_id":346,"authors":347,"venue":313,"cited_by_count":102,"oa_url":8,"card":8,"direction":204,"ingested_from":161},"W7221746877",[348,351,354,356],{"name":349,"orcid":350},"Alperen Kaan Bütüner","https:\u002F\u002Forcid.org\u002F0000-0002-2121-3529",{"name":352,"orcid":353},"Hilal Erdoğan","https:\u002F\u002Forcid.org\u002F0000-0002-0387-2600",{"name":355,"orcid":8},"İsmail Alper Susurluk",{"name":357,"orcid":358},"Edwin E. Lewis","https:\u002F\u002Forcid.org\u002F0000-0002-8825-1562","2026-10-16T23:30:09.556034Z"]