Plant phenotyping methods.

Vision Transformers Enable Advanced Plant Phenotyping in Controlled Environments →

Reliable plant segmentation in high-throughput phenotyping must transfer across species and imaging conditions without repeated model tuning or extensive reannotation. We compare three segmentation strategies using images from Oak Ridge National Laboratory's Advanced Plant Phenotyping Laboratory: (i) fixed color-based thresholding, (ii) supervised U-Nets trained from scratch, and (iii) pretrained vision transformers fine-tuned for binary segmentation. Models were evaluated on a held-out test set and a generalization set that comprised unseen species. On the held-out test set, thresholding, the best U-Net, and the best vision transformer achieved mean Dice scores of 58.3, 96.6, and 97.3, respectively. On the generalization set, the corresponding Dice scores were 56.5, 86.2, and 95.7. Thresholding remained effective on some datasets but failed when plant appearance changed. Supervised U-Net training resolved within-distribution errors but failed to generalize to novel species and backgrounds. Pretrained vision transformers consistently produced high-accuracy segmentations across the evaluated species, views, soil backgrounds, and tray types. These results benchmark the practical progression from fixed rules to task-specific supervision and pretrained visual representations for controlled-environment plant phenotyping.

Why it matches plant phenotyping methods植物フェノタイピングにおける画像セグメンテーション手法を比較・検証し、異なる種や撮像条件への汎化性能をベンチマークしているため、方法が研究の中心である。

abstractReliable plant segmentation in high-throughput phenotyping must transfer across species and imaging conditions without repeated model tuning or extensive reannotation.

MC-SlotNet: Multiplicity-Consistent Slot-Based Full-Cell Instance Segmentation for Overlapping Plant Suspension-Culture Microscopy →

Overlapping cells in plant suspension-culture microscopy pose a particular challenge, for instance, segmentation because a single pixel may belong to more than one cell. Most standard instance-segmentation methods are not designed for this setting and tend to treat overlapping objects as mutually exclusive regions. We instead represent each cell as an independent full-cell instance and introduce MC-SlotNet, an architecture that separates competitive object-slot feature assignment from mask decoding. This allows multiple predicted masks to occupy the same image region. We further introduce a mask-level multiplicity-consistency loss that encourages the predicted number of masks covering a pixel to agree with the underlying cell occupancy. We evaluate MC-SlotNet on a newly annotated dataset of 53 Siraitia grosvenorii suspension-culture micrographs containing 4131 full-cell instances acquired at 4×–40× magnification. Using grouped five-fold cross-validation and an overlap-preserving evaluation protocol, we compare the method with Mask R-CNN, SOLOv2, and Mask2Former. MC-SlotNet achieves the best performance on AP50 (0.800), mAP50:95 (0.565), F150 (0.842), all-ground-truth Dice (0.761), AJI+ (0.754), overlap-region Dice (0.705), and overlap-instance recall (0.828). Its AP75 (0.649) is comparable to Mask2Former’s (0.651). MC-SlotNet also has the lowest inference time among the evaluated methods, at 1.113 s/image. These results indicate that decoding full-cell masks independently, rather than enforcing an exclusive partition of image pixels, is well-suited to instance segmentation in plant suspension-culture microscopy images with substantial cell overlap.

Why it matches plant phenotyping methods植物細胞の重複画像から個々の細胞を分割・抽出する新規モデルを開発し、データセット上で既存手法と比較評価しており、植物表現型取得法が研究の中心である。

abstractWe instead represent each cell as an independent full-cell instance and introduce MC-SlotNet, an architecture that separates competitive object-slot feature assignment from mask decoding.

Deep convolutional neural network-based automated identification and classification of mungbean foliar diseases →

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.

Why it matches plant phenotyping methods圃場画像からマングビーン葉の病徴・病害状態を推定する深層学習手法を複数モデルで評価・検証しており、植物病害フェノタイピング手法が中心です。

abstractsystematic multi-architecture evaluation for mungbean disease classification under natural field conditions remains limited

From occlusion to 3D: amodal completion-assisted single-view wheat reconstruction →

Occlusion is a major factor limiting accurate three-dimensional (3D) wheat phenotyping. In natural growth conditions, overlapping spikes, leaves, and stems often make only partial target regions visible in single-view images, hindering complete and reliable 3D reconstruction. To address this problem, this study proposes an amodal completion-assisted, sequential framework for single-view 3D reconstruction of occluded wheat. The framework first uses visible prompts to recover the complete appearance and structural cues of occluded targets, and then feeds the completed images into single-view 3D reconstruction models to generate complete 3D structures. We construct the MMWO (Multi-view Multi-instance Wheat Occlusion) dataset from MMW, which is captured under controlled indoor scenarios, by synthesizing diverse occlusion samples through organ-level cutouts, random geometric transformations, and region-constrained pasting, with annotations including visible masks, occlusion masks, and complete target images. Six representative reconstruction methods, including Direct3D, Real3D, SF3D, Spar3D, TRELLIS.2, and Hunyuan3D, are systematically evaluated. Hunyuan3D achieves the best geometric performance, with the lowest mean CD- \(L_1\) and CD- \(L_2\) values of 0.1286 and 0.0536, and the highest mean F-score of 0.5668. SF3D achieves the best rendering quality in terms of PSNR, SSIM, and LPIPS. In addition, Pix2Gestalt completion reduces the estimation errors of spike length, width, and area from 9.31%, 10.89%, and 32.23% to 4.64%, 9.70%, and 9.45%, respectively. These results demonstrate that amodal completion can effectively alleviate occlusion-induced information loss and provide more complete structural priors for single-view 3D wheat reconstruction. This study offers a feasible solution for 3D wheat phenotyping under occlusion and provides a systematic reference for applying 3D generative models to agricultural phenotyping.

Why it matches plant phenotyping methods遮蔽下の単一画像から小麦器官を3D再構成し、形質推定精度を改善する手法を開発・比較検証しており、植物フェノタイピング手法が中心である。

abstractTo address this problem, this study proposes an amodal completion-assisted, sequential framework for single-view 3D reconstruction of occluded wheat.

Deep spatial-spectral fusion of UAV RGB and hyperspectral imagery for potato plant disease detection →

Accurate plant disease detection remains challenging when using single-modality data, which fails to capture comprehensive disease-related features. However, many existing studies rely on pixel-level classification or prior plant segmentation and lack explicit modeling of cross-modal interactions, limiting their ability to distinguish between healthy and diseased plants. This study presents a spatial–spectral fusion deep learning framework (S2-PDD) that fuses uncrewed aerial vehicles (UAV)-based RGB imagery and hyperspectral-derived feature representations for plant-level detection of potato diseases (blackleg and Potato virus Y (PVY)), employing early fusion (modalities combined at the input stage) and middle fusion (features integrated at intermediate stages within the model backbone) strategies. The multimodal fusion models were compared against single-modal models and an existing S2ADet model. Model performance, assessed through five-fold cross-validation, demonstrated that multimodal models integrating RGB and vegetation index features achieved the highest mAPs of 86.65 ± 1.71 (%; E-RV model) and 85.74 ± 1.96 (M-RV), respectively. These mAPs were higher than those of all single-modal models, including the RGB-only (83.21 ± 1.46) and hyperspectral-only models (PCA features: 79.71 ± 1.45; vegetation index features: 85.31 ± 2.36). They also exceeded mAPs of multimodal models combining RGB with PCA features (early fusion: 83.00 ± 2.81; middle fusion: 83.11 ± 2.46; S2ADet: 84.04 ± 2.73), regardless of the fusion strategy. The superior performance highlights that vegetation index features provide strong class separability compared to other hyperspectral representations. The proposed models achieved strong plant-level detection performance, with AP of 78.65 ± 4.14 (E-RV) and 77.40 ± 3.46 (M-RV) for blackleg disease, as well as 84.82 ± 3.59 (E-RV) and 83.28 ± 3.86 (M-RV) for PVY. These results demonstrate the potential of UAV-based multimodal sensing for disease monitoring in cropping systems. A potato plant disease detection dataset was constructed and made publicly available, containing paired RGB and hyperspectral image tiles with bounding box annotations. The code is available at https://github.com/Tim-Agro/S2-PDD.

Why it matches plant phenotyping methodsUAV RGB・ハイパースペクトル画像からジャガイモ個体の病害状態を推定する融合モデルを開発・比較検証し、公開データセットも構築しており、病害表現型の取得・抽出手法が中心である。

abstractThis study presents a spatial–spectral fusion deep learning framework (S2-PDD) that fuses uncrewed aerial vehicles (UAV)-based RGB imagery and hyperspectral-derived feature representations for plant-level detection of potato diseases (blackleg and Potato virus Y (PVY))

iPheno: A Novel Dual-Aware Vision-Language Model for Multi-Task Fine-Scale Crop Phenotyping. →

Crop phenotyping is crucial for advancing plant breeding, yet remains a significant challenge. Manual approaches are labor-intensive and do not scale to the analysis of large datasets, while computational methods like Vision-Language Models (VLMs) lack the adaptability for fine-scale spatial reasoning and diverse phenotyping scenarios. To bridge the gaps, we present iPheno, a fully open-source, domain-specialized multimodal VLM for fine-scale crop phenotyping. A key innovation of iPheno is its dual-aware architecture. First, a spatial-aware feature extractor samples mask regions into local K-Nearest Neighbors (KNN) graphs and enables fine-scale analysis of arbitrary-shaped regions; second, a task-aware Mixture-of-Experts (MoE) routing mechanism activates specialized modules for each phenotyping task. To train iPheno and achieve rigorous benchmarking, we constructed iPheno-120K, a large-scale high-precision dataset designed for multiple phenotyping tasks. Evaluations on iPheno-120k test set and other publicly available datasets showed that iPheno outperformed all fine-tuned baselines, by improving F1-score by 17.1% (LLaVA-1.6-13B) to 28.9% (MiniCPM-o-9B), while achieving the highest inference speed and memory efficiency. A web server (https://ipheno.ai4bread.com), a mobile application (www.ipheno.cn), and a stand-alone PC client (https://github.com/2997029323/iPheno-PC-Client) are available for iPheno.

Why it matches plant phenotyping methods植物表現型を対象とするVLMの開発、マルチタスク評価、専用データセット構築が研究の中心であり、明確な方法論的貢献がある。

abstractwe present iPheno, a fully open-source, domain-specialized multimodal VLM for fine-scale crop phenotyping.

What does 'control' mean in plant pathology? A systematic review of measurement and analysis. →

Abstract Disease management aims to protect crop yield and quality and reduce economic losses caused by plant pathogens. Consequently, reducing disease is a central objective of applied plant pathology. However, what constitutes effective disease control, and how it is measured and analyzed, varies substantially among studies. We conducted a systematic methodological review to characterize how plant disease control has been evaluated in the plant pathology literature over the past 15 years. We searched selected plant pathology journals for articles containing "control" in their titles and used an artificial intelligence-assisted workflow, followed by human verification, to extract and standardize information on experimental settings, disease measurements, measurement scales, and statistical analyses. The final dataset comprised 340 articles representing diverse host-pathogen systems and experimental environments. Disease control was evaluated using a wide range of response variables, most commonly disease severity and incidence, with substantial heterogeneity in measurement scales and sampling practices. Despite this diversity, statistical analysis was remarkably uniform: 79.4% of articles relied exclusively on ANOVA-based approaches. Among studies using ordinal disease scales, 73.3% included ANOVA in the analysis, whereas only 10.6% explicitly reported data transformation. Mean-separation procedures were also common, particularly Tukey, Fisher's LSD, and Duncan's multiple range test; Duncan's test was reported in 20.3% of all articles and varied markedly among journals. Our findings reveal a marked contrast between diversity in how plant disease control is measured and the narrower range of methods used to analyze those measurements. Greater alignment among biological meaning, measurement properties, experimental design, and statistical analysis could improve transparency, comparability, and interpretation in disease-management research.

Why it matches plant phenotyping methods植物病害の重症度・発生率など、植物の病害状態をどのように測定・解析するかを体系的にレビューしており、測定尺度、サンプリング、統計手法の標準化が中心的な方法論的貢献である。

abstractWe conducted a systematic methodological review to characterize how plant disease control has been evaluated in the plant pathology literature over the past 15 years.

Spectral-Consistency-Aware Evaluation of Deep Super-Resolution Methods for UAV Five-Band Multispectral Crop Imagery →

Unmanned aerial vehicle (UAV)-based multispectral imaging enables flexible, non-destructive crop monitoring. Although UAV imagery offers much higher spatial resolution than satellite platforms, its effective spatial detail at typical operational flight altitudes can still be insufficient for plant-level interpretation and fine canopy structure, which can reduce vegetation-index reliability. Most super-resolution (SR) research targets RGB or satellite imagery and emphasizes perceptual or pixel-wise quality, leaving the spectral fidelity of reconstructed UAV multispectral imagery under-examined. This study benchmarked an SR evaluation framework for UAV-based five-band crop imagery (Blue, Green, Red, Red-edge, and near-infrared) using the open-source AI Hub cabbage dataset, with low-resolution inputs generated by controlled downsampling at ×2, ×3, and ×4. Nine methods (bicubic, SRCNN, EDSR, RCAN, SwinIR-based, ESRGAN-based, HAT-based, DAT-based, and DRCT-based SR) were compared under joint five-channel and band-wise reconstruction on 2170 test scenes using image-quality, spectral-angle, vegetation-index (NDVI, GNDVI, NDRE), band-wise, and efficiency metrics. EDSR and RCAN gave the most balanced performance. At ×4, band-wise reconstruction was strongest for per-band spatial fidelity, where EDSR reduced RMSE by 12.6%, and RCAN lowered near-infrared RMSE by about 21% relative to bicubic, whereas joint reconstruction with its spectral-angle and vegetation-index losses best preserved spectral relationships (spectral angle and vegetation-index errors). Learning-based gains were clearest at ×4. The recently proposed HAT-based, DAT-based, and DRCT-based attention models achieved the strongest pixel-wise RMSE and PSNR but did not surpass EDSR or RCAN on spectral angle or vegetation-index preservation under the equalized training budget. These results indicate that UAV multispectral SR should be assessed by spatial fidelity together with spectral consistency and agricultural index reliability.

Why it matches plant phenotyping methodsUAVマルチスペクトル画像の超解像法を比較・ベンチマークし、スペクトル一貫性や植生指数の信頼性を評価する研究であり、植物キャノピー形質の画像取得・抽出基盤が中心である。

abstractThis study benchmarked an SR evaluation framework for UAV-based five-band crop imagery

CED-RTDETR: a contour-aware evidence-guided decoupled network for rice leaf disease detection. →

To address the challenges posed by large lesion-scale variations, weak boundary cues, and high inter-class similarity in rice leaf disease detection, this study proposes an improved RT-DETRv2-R50-based rice disease detection model, termed CED-RTDETR. First, a Lightweight Contour-Guided Aggregation Backbone (LCGA-Backbone) is constructed. In the PResNet residual blocks, an InceptionDWConv2d-based direction-aware depthwise separable spatial mixing strategy is introduced to capture local, horizontal, and vertical lesion texture patterns with low computational overhead. Meanwhile, a Contour-guided Efficient Global Aggregation Block (CEGA Block) is embedded after the outputs of the C3, C4, and C5 stages. Through contour-difference enhancement, channel shuffle, group-wise efficient global aggregation, and bottleneck channel mixing, the proposed block strengthens the representation of the boundaries of small lesions, weak textures, and contextual semantics. Second, a Multi-scale Evidence Interaction Fusion Neck (MEIF-Neck) is used to perform multi-scale evidence interaction and salient-region competition after cross-scale feature concatenation. A lightweight feature reconstruction process is further implemented using Spatial-Evidence RepNCSPELAN (SE-RepNCSPELAN), which is developed from a YOLO-style feature fusion structure. Finally, a Direction-Amplitude Decoupled Deformable Attention (DAD-DA) mechanism is introduced to decompose sampling offsets into direction rotation residuals and radius gains while incorporating an aspect-ratio-aware geometric compression-restoration strategy, thereby improving the geometric stability of sampling locations during the decoding stage. Experimental results on the constructed rice leaf disease dataset show that CED-RTDETR achieves AP, AP50, and AP75 values of 29.5%, 75.2%, and 17.6%, respectively, outperforming RT-DETRv2-R50 by 4.1, 4.9, and 3.8 percentage points. To further evaluate the model on an additional public benchmark dataset, experiments were also conducted on the public Rice Disease Dataset. On this dataset, CED-RTDETR achieves AP, AP50, and AP75 values of 34.1%, 77.1%, and 23.2%, respectively, improving upon RT-DETRv2-R50 by 4.3, 5.6, and 3.4 percentage points. These results indicate that the proposed method achieves consistent overall performance improvements on both the constructed dataset and the public benchmark dataset.

Why it matches plant phenotyping methodsイネ葉の病斑・病害を画像から検出する深層学習モデルを開発し、構築データセットと公開ベンチマークで性能検証しているため、植物病害状態のフェノタイピング手法が中心です。

abstractTo address the challenges posed by large lesion-scale variations, weak boundary cues, and high inter-class similarity in rice leaf disease detection, this study proposes an improved RT-DETRv2-R50-based rice disease detection model, termed CED-RTDETR.

A multi-scale dataset combining 3D plant architecture, leaf gas exchange, and whole-plant fluxes in young oil palm under controlled climate scenarios →

Functional-structural plant models simulate plant responses to environmental conditions, but their development and evaluation are often limited by the lack of datasets combining detailed architectural and physiological measurements. Here, we present a comprehensive dataset acquired from four oil palm plants ( Elaeis guinnensis) grown under controlled and contrasting climate scenarios. The dataset includes (i) three-dimensional reconstructions of plant architecture derived from terrestrial lidar point clouds, (ii) leaf-level gas exchange measurements used to parameterize photosynthesis and stomatal conductance models, and (iii) continuous plant-scale measurements of CO 2 and H 2 O fluxes obtained in a microcosm under precisely monitored and manipulated environmental conditions (light, temperature, humidity, and CO 2 concentration) across height climate scenarios. By combining detailed structural data with physiological measurements at both leaf and whole-plant scales, this database has been designed to build and evaluate digital twins (or shadows) of plants functioning under controlled conditions. It provides a valuable resource for calibrating biophysical models (light interception and photosynthesis), benchmarking model predictions across scales, and investigating the consistency between leaf-level parameterization and plant-level fluxes. All data and processing workflows are openly available, facilitating reuse for model development, evaluation, and intercomparison in plant and crop modelling communities.

Why it matches plant phenotyping methods3D LiDARによる植物構造計測と生理計測を統合したデータセットで、モデルの較正・ベンチマーク・評価を主目的としており、植物フェノタイピング手法と再利用可能なワークフローが中心である。

abstractthree-dimensional reconstructions of plant architecture derived from terrestrial lidar point clouds

UAV-Hyp: UAV Dataset for Remote Sensing-Based Phenotyping of Hypericum perforatum →

Quantitative flowering phenotypes are needed to support breeding and harvest management in Hypericum perforatum L. (St. John’s wort), but manual flower assessment is slow and difficult to standardize under field conditions. UAV-Hyp is a multi-temporal UAV RGB dataset containing 12,653 high-resolution images acquired at 26 measurement dates across the complete flowering period of 15 H. perforatum accessions. The images represent variable illumination, soil moisture, weed pressure, and developmental stages. The dataset provides 59,163 plant bounding boxes and 107,054 flower bounding boxes. As an application example, cascaded YOLOv8 plant and flower detectors achieved mAP@0.50:0.95 values of 0.977 and 0.950, respectively. UAV-Hyp supports scalable flower quantification and the development of time-series phenotyping methods for genotype comparison and quality-oriented medicinal-plant breeding.

Why it matches plant phenotyping methods植物の開花形質を定量化するUAV画像データセットを提供し、検出性能も評価しているため、フェノタイピング用データセット・解析手法が中心です。

abstractUAV-Hyp is a multi-temporal UAV RGB dataset containing 12,653 high-resolution images acquired at 26 measurement dates across the complete flowering period of 15 H. perforatum accessions.

BFMambaNet: Boundary-Frequency-Guided Global Semantic Mamba Network for Fine-Grained Camellia oleifera Leaf Disease Segmentation. →

Camellia oleifera leaf disease segmentation under natural field conditions is important for precision plant protection but remains challenging because lesions often show small target areas, blurred boundaries, uneven illumination, complex backgrounds, and coexisting symptoms. To address these problems, this paper proposes BFMambaNet, a Boundary-Frequency-guided Global Semantic Mamba Network for fine-grained disease segmentation. The model adopts an encoder-decoder framework and introduces a Global Semantic Mamba-based spatial selective feature modeling block to capture long-range lesion context and reduce semantic confusion. A gated wavelet spatial enhancement block is further designed to strengthen high-frequency boundary details while suppressing noisy responses. During training, boundary-frequency auxiliary supervision guides contour localization and pathological texture recovery without additional manual boundary labels. A reinforcement-learning-guided adaptive loss controller adjusts class-wise reweighting factors and loss-component weights according to the training state, improving optimization stability. A pixel-level dataset containing 1400 images and seven disease categories was constructed for evaluation. Experimental results show that BFMambaNet achieves 92.39% Precision, 91.43% Recall, 91.26% Dice, and 85.46% mIoU, outperforming representative CNN-based, Transformer-based, and Mamba-based models. Evaluations on environmental subsets confirm superior robustness, outperforming VMamba by 3.70% mIoU under uneven illumination, 3.55% mIoU under complex backgrounds, and 5.10% mIoU under coexisting symptoms. Cross-dataset validation on Apple leaf diseases further proves its generalization with 3.39% mIoU and 3.84% Dice improvements over U-Mamba, while maintaining a competitive inference speed of 30 FPS. Qualitative results also show clearer boundaries, fewer missed small lesions, and more stable predictions in complex field scenarios.

Why it matches plant phenotyping methods植物葉の病斑・病害状態を画像から画素単位で推定するセグメンテーション手法を開発し、データセット構築、性能評価、環境別評価、クロスデータセット検証まで行っており、病害フェノタイピング手法が中心である。

abstractA pixel-level dataset containing 1400 images and seven disease categories was constructed for evaluation.

A Method for Measuring Plant Spacing of Maize Seedlings Based on Improved YOLOv8 →

The uniformity of maize plant spacing serves as a critical indicator for assessing sowing quality, seed vigor, and field seedling emergence stability. However, manual measurement is inefficient, and complex field conditions make automatic seedling detection and plant spacing measurement challenging. Aiming at the challenges of missed detection, insufficient accuracy for small targets, and large errors in plant spacing calculation under complex field conditions, this study constructs a high-quality dataset containing 693 maize seedling images and implements preprocessing enhancement for images degraded by haze or dust. An intelligent maize seedling detection and plant spacing measurement method based on improved YOLOv8 is proposed. The Global Attention Mechanism (GAM) is embedded into the backbone network to strengthen cross-dimension information interaction between channels and spaces, suppress background interference, and reduce the missed detection rate. The Bi-directional Feature Pyramid Network (BiFPN) is adopted to replace the original PAFPN for enhanced multi-scale feature fusion and deep semantic representation. A new 160 × 160 high-resolution small-object detection layer is added to significantly improve the detection performance of weak and small seedlings. Experimental results demonstrate that the improved model achieves a precision, recall, mAP50, and mAP50-95 of 89.4%, 90.3%, 94.4%, and 49.4%, respectively, which are 2.6, 0.5, 1.5, and 2.7 percentage points higher than those of the original YOLOv8 model. These results indicate that the proposed model improved maize seedling detection performance under complex field conditions. Automatic plant spacing calculation is realized based on detection outputs; the average plant spacing of the dataset is 30.42 cm, with a relative error of only 4.93% compared with the preset sowing spacing of 32 cm. The proposed method can efficiently accomplish field seedling identification, plant spacing quantification, and sowing quality evaluation, providing reliable technical support for precision maize sowing, seeder parameter optimization, and intelligent field management, which is of great significance for promoting the intelligent upgrading of grain crop production.

Why it matches plant phenotyping methodsトウモロコシ幼苗の検出画像から株間という植物形態・配置形質を自動抽出するYOLOv8手法を開発し、データセット構築と性能検証まで行っており、フェノタイピング手法が中心である。

abstractAutomatic plant spacing calculation is realized based on detection outputs

Precision Greenhouse Rose Phenotyping from UAV Imagery Using a Multi-Source Dataset and Lightweight BloomRoseNet →

Accurate detection of blooming roses and flower buds is essential for greenhouse phenotyping, cultivation scheduling, harvest planning, and yield management. However, UAV-derived greenhouse imagery presents major challenges because rose targets are often small, densely distributed, partially occluded, and visually similar to complex backgrounds. This study proposes a lightweight rose detection framework that combines multi-source dataset construction with an improved YOLOv12n-based detector, termed BloomRoseNet. A GreenHouse Rose dataset was constructed by integrating self-collected UAV overhead images, screened RoseTracker images, and supplementary multi-view rose images to increase diversity in scale, growth stage, viewpoint, and background complexity. BloomRoseNet introduces task-oriented improvements for fine-grained feature extraction, adaptive feature fusion, and attention-enhanced detection. The supplementary multi-view data improved precision, recall, and mAP@50 from 85.2%, 82.8%, and 89.2% to 86.1%, 85.3%, and 90.5%, respectively. Compared with the baseline YOLOv12n, BloomRoseNet increased precision, recall, mAP@50, and mAP@50:95 by 3.2, 3.3, 3.6, and 1.6 percentage points, respectively, while reducing parameters from 2.55 M to 2.08 M and model size from 5.5 MB to 4.5 MB. The model also maintained real-time inference capability and stronger robustness under blur, occlusion, and illumination disturbances. The proposed framework provides an effective and practical solution for UAV-based greenhouse rose monitoring and supports precision cultivation management.

Why it matches plant phenotyping methodsバラの開花・蕾を対象としたUAV画像フェノタイピング手法を開発し、データセット構築、検出モデル改良、性能評価を中心に扱っているため。

abstractThis study proposes a lightweight rose detection framework that combines multi-source dataset construction with an improved YOLOv12n-based detector, termed BloomRoseNet.

Deep learning enables edge deployment for citrus leaf disease recognition in natural orchard scenes. →

Background Accurate and efficient detection of citrus leaf diseases is important for orchard monitoring, early intervention, and intelligent disease management. However, practical application in natural orchard environments remains challenging because of complex backgrounds, large variation in symptom scale, strong interclass similarity, and limited edge computing resources. Methods In this study, a lightweight object detection model, SKM-YOLOv11, was developed based on YOLOv11n for citrus leaf disease recognition in natural scenes. A self-built dataset was constructed from public images and field images, containing 6414 images across six classes. StarNet-S050 was introduced as the backbone, C3k2-Star was designed to enhance feature fusion across scales, and a lightweight shared detection head, MNS-Head, was constructed to reduce prediction redundancy. Results Compared with YOLOv11n, SKM-YOLOv11 increased Precision, Recall, and mAP @0.5 by 2.92, 1.47, and 0.95 percentage points, respectively. Meanwhile, FLOPs, parameter count, and model size were reduced by 31.7%, 32.7%, and 34.55%, respectively. Edge deployment on Jetson Orin NX Super achieved an inference speed of 72.45 frames per second. Conclusion The proposed model has strong potential for real-time citrus disease screening on resource-constrained devices and provides a feasible solution for edge-based intelligent disease management in natural orchards.

Why it matches plant phenotyping methods柑橘葉の病害状態を画像から認識する軽量物体検出モデルを開発し、データセット、精度比較、エッジ実装性能まで評価しており、植物表現型（病害状態）の取得・推定手法が中心である。

abstracta lightweight object detection model, SKM-YOLOv11, was developed based on YOLOv11n for citrus leaf disease recognition in natural scenes.

Depth4PH: a vision foundation model-based framework for plant height estimation in agricultural scenes →

Plant height is a key 3D phenotypic trait for assessing crop growth, biomass accumulation, and lodging resistance. To overcome the practical limitations of conventional plant height measurement methods, this study proposes a novel vision foundation model-based framework named Depth for Plant Height (Depth4PH) for plant height estimation in agricultural scenes using low-cost monocular RGB imaging. As part of our contributions, a synthetic–real coupled multimodal dataset was constructed by integrating Blender virtual agricultural scenes (Blender VAS) with real field images. Building upon the existing Depth Anything V2 foundation model, we developed a novel module called Transfer-based Agricultural Metric Depth Anything V2 (TAM-Depth V2) for absolute metric depth estimation through parameter-efficient fine-tuning, depth decoder reconstruction, and joint loss optimization. Furthermore, we designed a novel multi-source prompt-based segmentation framework, MSP-SAM2, to generate positive and negative prompts for zero-shot crop instance segmentation. Finally, a new inverse physical plant height estimation algorithm, RANSAC-Per, was introduced to estimate plant height by combining truncated percentile statistics with local RANSAC micro-plane fitting, thereby reducing the effects of depth noise and field microtopographic variation. The result showed that TAM-Depth V2 achieved stable absolute depth estimation, with an RMSE of 0.1162 m and an AbsRel of 4.25%. Compared to the original box-prompted SAM 2, MSP-SAM2 achieved a 4.4% improvement in mIoU, reaching 91.6% and a recall of 93.2%. On a 350-plant multi-crop test set, Depth4PH achieved R 2 = 0.948, RMSE = 12.23 cm, and MAE = 8.82 cm, and MAPE =10.15%, with crop-specific RMSEs ranging from 3.99 cm (cucumber) to 20.73 cm (maize), significantly outperforming the traditional Global-MinMax baseline (which had an RMSE of 22.62 cm). These results indicate that Depth4PH provides a promising foundational pathway for high-throughput crop phenotyping. With future optimization for edge deployment, it holds significant potential to support high-throughput monitoring in precision agriculture.

Why it matches plant phenotyping methods植物高の画像取得・深度推定・セグメンテーション・高さ抽出アルゴリズムを一体化した植物表現型計測フレームワークの開発と検証が中心であり、データセット構築と性能評価も含む。

abstractthis study proposes a novel vision foundation model-based framework named Depth for Plant Height (Depth4PH) for plant height estimation in agricultural scenes using low-cost monocular RGB imaging.

HagPF: Hierarchical-annotation-guided phenotypic framework for stem instance segmentation and length measurement in plant point clouds →

As a critical structural component that connects almost all other types of plant organs, the stem system not only supports the weight of the total plant, but also serves as a vital channel for nutriment transportation. Accurate phenotypic measurement of stem instances is of practical significance for assessing crop growth dynamics and predicting yield. To address current 3D phenotyping challenges of crops such as the difficulty in separating stem segments from the stem system and the low accuracy in stem length measurement, we propose a Hierarchical-annotation-guided Phenotypic Framework (HagPF) for Stem Instance Segmentation and Length Measurement in plant point clouds. Specifically, a hierarchical leaf-stem organ instance annotation strategy is devised to effectively train a PSegNet network for leaf and stem instance segmentation. The segmentation is then followed by a shape-adaptive measurement algorithm to automatically measure the length of stem segments that are morphologically diverse in space. On a 3D dataset comprising four crop species, the proposed framework achieved an Intersection over Union (IoU) of 95.45% for organ semantic segmentation and a Mean Weighted Coverage (mWCov) of 87.87% for instance segmentation (both stem and leaf). Regarding to the stem length measurement, the method obtained an average Root Mean Square Error (RMSE) of 1.044 cm and a relative error of 11.907%, outperforming 7 mainstream methods. The relevant dataset and source code can be found at: https://github.com/Jinx00/stem-length-measurement .

Why it matches plant phenotyping methods植物点群から茎のインスタンスを分割し、茎長を自動測定する3D表現型解析フレームワークの開発・比較検証が中心である。

abstractwe propose a Hierarchical-annotation-guided Phenotypic Framework (HagPF) for Stem Instance Segmentation and Length Measurement in plant point clouds.

Field-scale crop growth stage mapping using multispectral images and deep hierarchical segmentation →

Accurate, field-scale mapping of crop growth stages is critical for supply-sensitive vegetable production, where timely harvests require detailed phenological information. Consecutive growth stages often involve rapid and subtle morphological changes and are influenced by challenging open-field conditions, which frequently result in misclassification when stages are treated as independent, discrete categories. To address this issue, CropMap is proposed as a growth-stage mapping framework that integrates Hierarchical Semantic Segmentation Networks (HSSN) with multispectral unmanned aerial vehicles (UAVs) imagery. CropMap incorporates the structured biological progression of crop development into the learning objective through tree-based label constraints, allowing the model to recognize phenological continuity and reduce confusion between adjacent stages. The framework is evaluated on the publicly available National Information Society Agency of Korea (NIA) field crop growth-stage dataset, a large-scale, multi-institutional UAV dataset containing 337,665 multispectral patches across six hierarchically related growth stages of Chinese cabbage and radish, curated by the NIA. CropMap achieves a test-set mean Intersection over Union (mIoU) of 0.5382, representing a 5% relative improvement over the best-performing transformer baseline (SegFormer; mIoU = 0.5124). Performance varies across classes: background separation is strong (IoU = 0.9128) and the rosette stage is well distinguished (IoU = 0.6354), while the leaf expansion stage remains the primary challenge (IoU = 0.3541), reflecting the inherent difficulty of mapping this spectrally and morphologically transitional class. These findings indicate that hierarchy-aware learning reduces inter-stage confusion for most phenological classes, but transitional growth stages remain a significant limitation for field-scale deployment. The framework provides a foundation for stage-resolved crop monitoring to support harvest timing and supply forecasting in high-value vegetable systems.

Why it matches plant phenotyping methods作物の生育段階という植物状態を、マルチスペクトルUAV画像と階層型セマンティックセグメンテーションで推定する手法が研究の中心であり、公開データセット上で性能評価も行っている。

abstractCropMap is proposed as a growth-stage mapping framework that integrates Hierarchical Semantic Segmentation Networks (HSSN) with multispectral unmanned aerial vehicles (UAVs) imagery.

A high-resolution (500 m) dataset for mapping key agronomic growth stages of maize in Northeast China. →

Accurate monitoring of agronomic phenology is essential for yield estimation and food security assessment. However, currently available maize phenology datasets usually represent only a limited number of growth stages, restricting their application in process-based crop modeling and stage-specific agricultural management. Here, we present a high-resolution maize phenology dataset for Northeast China spanning 2001-2024 at 500-m spatial resolution and daily temporal resolution. By coupling MODIS spectral information with meteorological drivers in an energy-driven XGBoost framework, we retrieved eight key agronomic stages: Emergence, Three-leaf, Seven-leaf, Jointing, Flowering, Silking, Milking, and Maturity. Validation against observations from 91 agrometeorological stations during 2009-2024 demonstrates robust performance, with an overall RMSE of less than 5 days and R² values greater than 0.63 across all stages. Beyond overall accuracy, the dataset shows strong spatial consistency and temporal stability, preserves coherent regional phenological gradients, and captures interannual variations over the 24-year period. This long-term, multi-stage dataset provides a valuable benchmark for crop model calibration, climate change impact assessment, and the development of adaptive agricultural strategies in one of the world's major maize-producing regions.

Why it matches plant phenotyping methodsMODISスペクトル情報と気象データ、XGBoostを組み合わせてトウモロコシの8つの生育段階を推定し、観測データで検証した高解像度フェノロジーデータセットであり、植物形質の取得・抽出手法とベンチマークが中心です。

abstractHere, we present a high-resolution maize phenology dataset for Northeast China spanning 2001-2024 at 500-m spatial resolution and daily temporal resolution.

A multimodal geospatial foundation model anticipates crop stress and yield failure across climates and species →

Introduction Climate extremes increasingly threaten agricultural production, yet many artificial intelligence systems in agriculture remain local, reactive and narrowly trained for one crop, region or sensing modality. Methods We present AgriFM, a multimodal geospatial foundation model that combines satellite image time series, radar, thermal observations, weather trajectories, soil properties, topography and sparse management variables to estimate crop-stress and yield-failure risk across crops and regions. AgriFM was pretrained using self-supervised objectives on 2.4 million field-season sequences and evaluated on a curated benchmark spanning maize, wheat, soybean, rice and sorghum across five agroclimatic regions. Results In held-out geography and time-split evaluations, AgriFM improved early stress detection and yield-failure prediction over statistical, crop-model and deep-learning baselines. The largest gains occurred during compound drought and heat events, for which AgriFM produced alerts 18 to 24 days earlier than the satellite-only baseline while maintaining improved calibration. Phenology-conditioned fusion improved transfer across planting calendars, and uncertainty calibration reduced false alerts at fixed recall. Discussion Because the study is based on retrospective datasets, these findings establish cross-region retrospective performance rather than prospective field efficacy. The results support further field-based evaluation of multimodal foundation models for climate-resilient crop monitoring.

Why it matches plant phenotyping methods作物ストレス状態と収量失敗リスクを衛星・レーダー・熱画像等から推定する基盤モデルを開発し、複数作物・地域のベンチマークで評価しており、植物状態の取得・推定手法が中心である。

abstractWe present AgriFM, a multimodal geospatial foundation model that combines satellite image time series, radar, thermal observations, weather trajectories, soil properties, topography and sparse management variables to estimate crop-stress and yield-failure risk across crops and regions.

Early Multi-Class Disease Detection in Chili Plant Leaves Using Convolutional Neural Networks: A Comparative Study →

Abstract Chili is an important economic and nutritional crop with a relatively limited availability of different disease-resistant varieties. Leaf diseases, including those caused by fungi, bacteria, viruses, pests, or nutritional deficiencies, significantly compromise production and crop quality. Early detection of these diseases is key to reducing yield loss; however, traditional visual examinations are limited by time constraints, human subjectivity, and low detection sensitivity at early stages of infection. To overcome these challenges, this study proposes a deep learning–based framework for early multi-class detection of chili leaf diseases using convolutional neural networks (CNNs). A real-field dataset comprising 16,392 high-resolution images of chili leaves across six disease and healthy classes was collected from multiple regions of Bangladesh. The dataset was preprocessed, augmented, and split into training and validation sets using an 80:20 ratio. Five pre-trained CNN architectures—DenseNet121, EfficientNetB3, MobileNetV2, ResNet50, and InceptionV3 were evaluated using a transfer learning strategy. Experimental results demonstrate that MobileNetV2 achieved the best performance, attaining an overall classification accuracy of 96%. The results indicate that the proposed system demonstrates strong generalization capability and effectively discriminates visually similar chili leaf diseases. This work can be considered a valuable application in precision agriculture, providing an efficient, automated, and practical approach for in situ early diagnosis of chili leaf diseases through smart farm management.

Why it matches plant phenotyping methods唐辛子葉の病害状態を画像から分類するCNN手法の開発・比較評価とデータセット構築が研究の中心であり、植物病害フェノタイピングに該当する。

abstractA real-field dataset comprising 16,392 high-resolution images of chili leaves across six disease and healthy classes was collected from multiple regions of Bangladesh.

WheatDisease-HRY: A Real-Field Wheat Disease Dataset with Baseline Deep Learning Benchmarks for Automated Disease Detection →

In India, to increase the wheat yield production and support sustainable agricultural practices, timely and correct identification of wheat crop diseases such as yellow rust, brown rust, etc., is very much essential. Further, the majority of deep learning algorithms show high efficiency in plant disease detection and recognition, but these studies rely on handcrafted datasets captured under controlled laboratory conditions and accordingly will not perform well in the real world domain. Furthermore, there is a lack of real field wheat crop disease datasets collected in India, particularly from Haryana state, being the largest producer of wheat crops. Therefore, to address the issue of lack of region specific crop diseases dataset, which are region-specific, the wheat crop disease data set is curated, consisting of 2672 images of healthy and diseased leaves of wheat crops collected from the real field of CCS Haryana Agricultural University, Hisar. The dataset includes typically three classes of leaves of wheat crop, i.e., Yellow Rust (910), Brown Rust (922) and Healthy leaves (840) collected using both a DSLR camera and different smartphone cameras under natural lighting and field conditions. All images were validated and labeled by taking the expertise of wheat pathologists to ensure reliability on the dataset. The study also provides a comprehensive, systematic workflow for transforming raw data into a high-quality benchmark dataset for training using image preprocessing and augmentation techniques. Besides this, a comparative benchmarking analysis is performed under identical experimental conditions using three widely adopted deep learning architectures, which includes ResNet50, MobileNetV2 and EfficientNet-B0. The experimental results show that ResNet50 and EfficientNet-B0 achieve similar performance i.e. approximately 91% classification accuracy on real-field data. However, MobileNetV2 offers a lightweight alternative suitable for mobile and edge deployment. Furthermore, Grad-CAM based explainability analysis was performed to validate model predictions and highlight disease specific regions in wheat leaves. Therefore, this study contributes a practical region specific WheatDisease-HRY dataset and baseline benchmarking framework for developing robust AI based wheat disease diagnosis tools for real world agricultural applications in India.

Why it matches plant phenotyping methodsコムギ葉の病徴を画像から判定するデータセットを構築し、前処理・拡張、深層学習ベンチマーク、説明可能性解析までを中心的に扱うため、植物病害状態の画像ベース表現型手法として採用。

abstractThe study also provides a comprehensive, systematic workflow for transforming raw data into a high-quality benchmark dataset for training using image preprocessing and augmentation techniques.

Low-cost monocular RGB-based 3D structural mapping for horticultural plants via semantic scene completion →

Precision agriculture increasingly relies on detailed structural information, such as canopy height and canopy volume, to enhance crop health monitoring and operational safety. However, existing methods based on costly LiDAR or RGB-D sensors are often impractical for large-scale deployment in dynamic and unstructured horticultural environments. Furthermore, conventional 2D segmentation and SLAM-based pipelines typically generate sparse, geometrically inconsistent semantic maps which are insufficient for actionable structural analysis in agricultural applications. To overcome these limitations, we propose a monocular 3D structural mapping framework tailored for horticultural plants via semantic scene completion. At inference, the proposed model takes a single RGB image as input and predicts voxel-wise geometry and semantics, from which task-oriented structural maps, including canopy height, canopy volume, and obstacle-aware traversability layers, are derived. Specifically, we first introduce a Depth-Aware Decoder Module that explicitly recovers depth in the spatial domain and fuses 2D-to-3D features, thereby mitigating depth ambiguity and reducing reliance on accurate pose. Second, an NCS-Guided Geometry Encoder is designed to inject normalized depth into voxel positional embeddings, enabling self-attention to perform global relational modeling within a depth-aware geometric coordinate system. In addition, a Global Encoder is utilized to refine local structural details, while an occupancy head produces the final 3D semantic completion outputs. We construct a horticultural 3D semantic scene dataset using an RGB-D sensor, which serves as a benchmark for evaluating our method, while the deployed model remains RGB-only. Extensive quantitative and qualitative experiments are conducted on both the Semantic-KITTI dataset and our dataset. On our dataset, the method achieves 82.31% occupancy IoU, 84.26% mIoU, and 86.25% precision. Beyond voxel-level evaluation, manual field measurements further show canopy height MAE values of 0.019-0.026 m and canopy volume proxy relative errors of 8.4%-11.4%. These results demonstrate the effectiveness of our approach in real-world agricultural scenarios, providing actionable structural insights for crop monitoring and autonomous robotic operations.

Why it matches plant phenotyping methods単眼RGB画像から植物の樹冠高・樹冠体積などの構造形質を推定する3Dフェノタイピング手法を開発し、データセット構築と実測検証も行っているため、方法が研究の中心である。

abstractwe propose a monocular 3D structural mapping framework tailored for horticultural plants via semantic scene completion.

An End-to-End Precision Phenotyping Framework: Rice Panicle Detection and Counting in Complex Fields via Lightweight DETR →

Accurate, high-throughput quantification of rice panicles is important for yield estimation and breeding-oriented rice phenotyping. However, unmanned aerial vehicle (UAV)-based panicle detection remains challenging because flight-altitude variation produces large target-scale changes. Flooded paddy backgrounds, leaf occlusion, and illumination fluctuations further obscure small panicle targets. To address these challenges, we constructed a composite multi-altitude dataset covering UAV imagery acquired from 3 to 20 m under varying field conditions. We then propose Panicle-DETR, a lightweight detection and counting framework based on a frequency-aware Cross Stage Partial (CSP) backbone. Rather than treating the Fast Fourier Transform (FFT) as a filter by itself, the proposed FasterFD module uses frequency-domain representations with learnable frequency-response reweighting to enhance panicle-related texture cues and reduce redundant background responses. A Lossless Feature Encoder is designed to preserve fine spatial information for small targets across altitude-induced scale changes, while a composite metric loss based on Normalized Gaussian Wasserstein Distance (NWD) and Inner-IoU improves localization for adherent and overlapping panicle clusters. On the composite dataset, Panicle-DETR achieved a Precision of 90.97%, a Mean Absolute Error of 4.28, and an R2 of 0.957 for single-frame panicle counting. With 13.78 M parameters and 53.0 GFLOPs, the framework achieved 16.9 FPS with 1.96 GB peak GPU memory in a battery-powered notebook benchmark, supporting its potential for resource-constrained field-side UAV image analysis.

Why it matches plant phenotyping methodsUAV画像からイネ穂の検出・計数という植物形質推定を中心に、マルチ高度データセットと専用解析フレームワークを開発・評価しているため。

abstractAccurate, high-throughput quantification of rice panicles is important for yield estimation and breeding-oriented rice phenotyping.

Multi-annual Multispectral Image Dataset of Chardonnay Grapevine Leaves with Yellowing disease and Easily Confused Symptoms →

Grapevine yellows diseases, including Flavescence Dorée (FD) and Bois Noir (BN), severely affect vineyards, making reliable detection methods essential. Detecting symptoms, particularly in Chardonnay variety, remains challenging due to year-to-year vine variability and the presence of symptoms that can be misleading. We present a dataset of multispectral images of grapevine leaves collected over four years (2021–2024). The dataset includes healthy leaves, grapevine yellows - overwhelmingly BN, and three other diseases (Esca, Discoloration, and Leafroll) with visually similar symptoms particularly for Leafroll, providing a diverse benchmark for disease recognition. Ground-truth labels were assigned based on field inspections conducted by Comité Champagne expert viticulturists who selected the vines before acquisitions and analyzed the images after acquisition. Images were acquired under laboratory conditions using a DJI Phantom 4 Multispectral camera from leaves collected in the Plumecoq Comité Champagne experimental vineyards. Leaves were collected on designated vines and placed on polystyrene boards on a 2x2 grid. One RGB image and five multispectral images corresponding to the Blue (B - 450 nm ± 16 nm), Green (G - 560 nm ± 16 nm), Red (R - 650 nm ± 16 nm), Red-Edge (RE - 730 nm ± 16 nm), and Near-Infrared (NIR - 840 nm ± 26 nm) bands were captured using the same acquisition system (camera-to-board distance of approximately 90 cm) over the four years.. For the vast majority, ambient light was used; for some acquisitions, indirect halogen spotlights were used. The 512x512 leaf images were manually cropped from the acquired original images with the same offsets between the bands; offsets were already present due to the parallax phenomenon. The dataset is organized into six folders (B, G, R, RE, NIR, and RGB). Each leaf is represented by one RGB image and five corresponding multispectral band images, enabling multimodal analysis and machine learning for grapevine disease detection. Each image filename encodes acquisition and annotation information as follows: Characters 1–2: acquisition year (21 = 2021, 22 = 2022, 23 = 2023, 24 = 2024). Character 3: image type/band (0 = RGB, 1 = B, 2 = G, 3 = R, 4 = RE, 5 = NIR). Character 4: class (J = Grapevine Yellows, T = Healthy, S = Esca, E = Leafroll, D = Discoloration). Character 5: illumination condition (0 = ambient light, G = additional left halogen spotlight, D = additional right halogen spotlight, 2 = additional both left and right halogen spotlightsl). Characters 6–9: image identifier, numbered sequentially from 0000. Character 10: relabelling status. After image acquisition, all leaves were independently reviewed by expert viticulturists on the recorded RGB images. A value of 1 indicates that the expert confirmed the original field label, 0 indicates that the original label was considered incorrect based on the information visible in the RGB image, and D indicates that the sample was reclassified as Discoloration. This dataset complements the "Multi-annual spectral data of Chardonnay grapevine leaves" dataset published on Recherche Data Gouv. While the previous dataset contains spectral measurements, the present dataset provides multispectral images acquired from the same grapevine leaves, enabling multimodal analyses that combine spectral signatures with image-based information. Note, however, that there is no bijection between the respective files.

Why it matches plant phenotyping methodsブドウ葉の病徴をマルチスペクトル画像で取得し、専門家ラベル付きの疾病認識用データセット／ベンチマークとして提供しており、植物状態の画像ベース表現型計測が中心である。

abstractWe present a dataset of multispectral images of grapevine leaves collected over four years (2021–2024).

Hyperspectral imaging and dynamic selective peak transformer for early-stage classification of lettuce heat responses. →

Introduction Early-stage detection and classification of lettuce heat responses are essential for non-destructive phenotyping, yet conventional assessment mainly relies on visible symptoms and manual observation. Methods This study constructed a lettuce hyperspectral dataset comprising heat-sensitive and heat-tolerant varieties under control and high-temperature treatments, and proposed the Dynamic Selective Peak Transformer (DSPformer). DSPformer integrates edge-enhanced feature extraction, dynamic multi-scale spatial-spectral representation, Peak-k selective attention, and a confusion-aware dynamic focal loss to enhance discriminative features while reducing spectral redundancy, class imbalance, and inter-class confusion. Results Under the patch-level evaluation protocol, DSPformer achieved 96.22% accuracy, 95.55% recall, 96.35% precision, and 95.95% F1-score, outperforming the compared CNN- and Transformer-based models. Day-wise evaluation showed that DSPformer reached 82.61% accuracy on Day 1 and 96.55% on Day 3, before visible heat-stress symptoms appeared on Day 6. Under a plant-level partition protocol, DSPformer maintained robust performance with 93.76 +/- 0.49% accuracy. Additional evaluation on the Indian Pines benchmark further demonstrated the applicability of DSPformer to general hyperspectral image classification. Discussion These findings suggest that hyperspectral imaging can capture heat-stress-sensitive information beyond visual phenotypes, and that DSPformer provides a promising framework for early, non-destructive lettuce heat-response screening and hyperspectral phenotyping-assisted breeding.

Why it matches plant phenotyping methodsレタスの熱ストレス応答を非破壊的に早期分類するため、ハイパースペクトル画像データセットと新規Transformer手法を開発・評価しており、植物表現型取得・抽出が中心である。

abstractMethods This study constructed a lettuce hyperspectral dataset comprising heat-sensitive and heat-tolerant varieties under control and high-temperature treatments, and proposed the Dynamic Selective Peak Transformer (DSPformer).

3D micro–X-ray tomography of lumen space in dried stem, branch, and outer tissues of Moringa oleifera: A structural case study →

Abstract Moringa oleifera is widely used in dry tropical and subtropical regions due to its rapid growth and high nutritional value, yet its internal tissue organization has primarily been described using two-dimensional anatomical approaches. Here, we present a three-dimensional micro–X-ray computed tomography (micro-XCT) characterization of lumen space in stem, branch, and outer tissues (bark region) from a single M. oleifera individual. Samples were oven-dried prior to imaging; therefore, the quantified void fraction represents apparent lumen/void space in dried material and should not be interpreted as in vivo porosity. Micro-XCT datasets were segmented to quantify cross-sectional lumen area distributions and apparent void fraction from pooled reconstructed slices. All data originate from a single individual; reported metrics represent structural descriptors of pooled cross-sections and not replicated biological measurements. The stem dataset exhibited a dense arrangement of small lumen features and a high number of segmented objects, consistent with a compact woody tissue organization in the scanned region. The branch dataset showed a larger proportion of void space and a strongly right-skewed size distribution with a minority of large lumen features. The outer tissue dataset displayed heterogeneous void space organization, which likely reflects a mixture of cell lumens, intercellular spaces, and drying-related cracks, and therefore is reported descriptively without assigning xylem-vessel identity. This study provides a conservative 3D structural dataset and an image-analysis workflow for quantifying lumen space in dried M. oleifera tissues, complementing published anatomical descriptions. The results highlight strong within-plant heterogeneity across tissue types and underscore the importance of sample preparation and histological validation when interpreting micro-XCT measurements in woody plants.

Why it matches plant phenotyping methods乾燥植物組織の3D micro-XCT画像から管腔・空隙の構造形質を抽出する画像解析ワークフローとデータセットが研究の中心であり、単なる生物学的測定ではない。

abstractMicro-XCT datasets were segmented to quantify cross-sectional lumen area distributions and apparent void fraction from pooled reconstructed slices.

BioHackEU25 report: Towards a Robust Validation Service for Data and Metadata in ARC RO-Crates →

Robust validation of both research data and its accompanying metadata is essential for ensuring adherence to FAIR principles. Current approaches often handle these aspects separately, hindering a holistic quality assessment. Building upon previous BioHackathon work establishing ARCs (Annotated Research Context) as RO-Crates (ARC RO-Crate), we aim to develop and demonstrate an integrated validation strategy for FAIR digital objects. It distinguishes between validating the metadata descriptor and the payload data files.For the metadata descriptor, validation will ensure structural and semantic compliance to the base RO-Crate specification and the ARC-ISA family of RO-Crate profiles, using and extending the RO-Crate validator tool.For the payload data files, validation targets the actual content, since data files often require domain-specific structural and value constraints, which requires explicit schema definitions. For this, we will integrate Frictionless for checking data content against community standards (e.g. MIAPPE, as demonstrated in the HORIZON project AGENT). Crucially, this project will also explore mechanisms for specifying expected data structures’ requirements within the ARC RO-Crate itself. This aims to provide a more self-contained description of data, investigating how such internal requirements can be linked to data validation frameworks, complementing the crate’s metadata validation.The overall goal is to provide a powerful, holistic validation mechanism for ARC RO-Crates, enhancing their reliability, trustworthiness, and FAIRness. A MIAPPE-compliant plant phenomics dataset will serve as a use case. This integrated validation approach aims to streamline quality control for researchers and will be packaged as a deployable microservice, offering broad applicability across diverse research workflows.

Why it matches plant phenotyping methodsARC RO-Crateのデータ・メタデータ検証サービスを開発する研究で、MIAPPE準拠の植物フェノミクスデータセットを具体的ユースケースとして扱う。植物表現型データの再利用可能な検証ツール／ワークフローが中心である。

abstractA MIAPPE-compliant plant phenomics dataset will serve as a use case.

UAV-Derived Multispectral Datasets and Index-Guided Segmentation for Maize Water Stress and Common Rust Detection Under Real Field Conditions →

The segmentation model achieved Mean IoU values of 0.7723 for Water Stress 2025, 0.9164 for Common Rust 2025, and 0.9531 on the benchmark dataset. The classifier achieved 99.54% accuracy for the five-class task; however, the improvement over the strongest baselines and the RGB + multispectral configuration was limited. Therefore, the classification component is not presented as a substantially superior classification-only model. Instead, it is interpreted as an exploratory multimodal analysis that quantifies the contribution and limitation of RGB, multispectral, Wavelet, and GLCM branches under the adopted UAV dataset protocol. For classification-only deployment, simpler alternatives such as DenseNet201 or the RGB + multispectral configuration may be more practical because they provide comparable accuracy with lower architectural or preprocessing complexity. Ablation, modality-controlled, and 21-run stability experiments showed reproducible segmentation results and clarified the behavior of the classification branches. RGB and multispectral branches mainly provided the peak classification accuracy, whereas Wavelet and GLCM branches mainly affected offline convergence rather than final accuracy. RGB, NDVI, and NDRE visualizations were also added for qualitative support. Since direct physiological ground measurements were not available for all samples, the masks are interpreted as adaptive index-guided labels rather than direct physiological ground truth. Overall, the main evidence of practical benefit is associated with UAV-based dataset construction, adaptive index-guided segmentation, and field-scale stress/disease mapping, while the classification experiments should be interpreted as modality-contribution and convergence analyses rather than proof of a practically superior complex classifier.

Why it matches plant phenotyping methodsUAVマルチスペクトル画像によるトウモロコシの水ストレス・病害状態のセグメンテーション、データセット構築、再現性評価が研究の中心であり、植物表現型の取得・抽出手法として実質的です。

abstractOverall, the main evidence of practical benefit is associated with UAV-based dataset construction, adaptive index-guided segmentation, and field-scale stress/disease mapping

AI-Based Mango Plant Disease Detection System →

Mango (Mangifera indica L.) is among the most important commercial fruits grown throughout the world in the tropical and subtropical areas. Even though mangoes are economically important, their cultivation is continuously threatened by a wide variety of leaf and fungi diseases, resulting in crop losses of up to 15-30% annually. [1]. Conventional disease identification depends heavily on expert visual inspection—a process that is inherently slow, subjective, and largely impractical for smallholder farmers operating in remote areas with limited access to agronomic specialists. This paper provides an end-to-end deep learning-based approach towards automatic mango leaf disease detection along with an Android application for real-time deployment in the field. This work is built upon MangoLeafBD [1], an openly accessible dataset that consists of a total of 4,000 images of RGB color space for seven different disease categories and one healthy class, with each category having 500 samples collected from four separate orchards in Bangladesh. Three types of transfer learning models including VGG16 [2], MobileNetV2 [3], and DenseNet121 [4] were considered after applying two-phase fine-tuning based on pre-trained ImageNet weights. For each of the three types of neural network models tested, a series of preprocessing steps consisting of bilinear resizing to size 224 x 224, channel-wise normalization, and image augmentation (rotation, zoom, brightness adjustment, horizontal flip, and shear) was used to increase model accuracy for diverse real-world images. The training model was deployed using TensorFlow Lite (TFLite), which allowed it to be run in offline mode on mid-end Android phones, without needing internet connectivity. The app lets farmers upload leaf images or take images and get a diagnosis of the leaf diseases, along with possible treatments for them.This work contributes a replicable pipeline linking state-of-the-art deep learning research with practical precision agriculture, particularly for rural communities that currently lack access to timely agronomic advisory services.

Why it matches plant phenotyping methodsマンゴー葉画像から病害状態を推定する深層学習パイプラインとモバイル実装が研究の中心であり、植物病害フェノタイピング手法に該当する。

abstractThis paper provides an end-to-end deep learning-based approach towards automatic mango leaf disease detection along with an Android application for real-time deployment in the field.

Automated Plant Leaf Disease Diagnosis using Deep Learning →

Plant varieties are essential for the survival of human beings and animals, as they act as an alternative source of food, fiber, fodder, and other raw materials for domestic needs and industries in any society. Early identification of plant leaf diseases is very important for keeping the health of crops intact. Crops being infected can affect the overall yield of crops, which may be detrimental to the earnings of farmers. With the emergence of artificial intelligence technology, it has become possible to deploy systems for quicker identification of illnesses. This work has been carried out for the prediction of plant diseases based on visual phenotypic manifestations, such as images of leaves. For this purpose, the dataset has been created after retrieving data from the PlantVillage dataset. The clinical reliability of different deep learning models of various representational capacities has been tested while using ImageNet pre-trained parameters. The experimental results show that MobileNetV2 achieves the highest accuracy of 96.15%, outperforming deep CNN (76.92%) and medium CNN (61.54%). The test accuracy and class-wise F1-scores for the CNN are observed to be substantially very high. The generalization ability and result of DCNN and MCNN are moderate and poor, respectively, as observed. Additionally, the proposed CNN only uses the disease-affected areas on the leaf, thus making the result more interpretable

Why it matches plant phenotyping methods葉の画像から病徴を推定する深層学習手法を開発・比較し、モデル性能と解釈性を評価しているため、植物表現型取得が中心である。

abstractThis work has been carried out for the prediction of plant diseases based on visual phenotypic manifestations, such as images of leaves.

GCP-YOLO: high-precision detection of tiny chili flowers in complex greenhouse scenes. →

Precise identification of chili flowers and buds is essential for enabling intelligent robotic pollination, continuous crop monitoring, and early yield prediction in protected horticulture. Reliable visual sensing remains challenging because chili targets are extremely small, densely distributed, and frequently occluded by foliage, while greenhouse environments introduce strong illumination variations and background reflections. These factors often lead to insufficient feature extraction and unstable detection accuracy in existing models, limiting their practical deployment in automated monitoring systems. To address these challenges, this study proposes GCP-YOLO, a lightweight yet high-performance detection framework built upon the YOLOv11n architecture. The model enhances small-target perception through three key improvements. First, a redesigned Generalized Feature Pyramid Network (GFPN) strengthens cross-scale feature interaction, improving the fusion of fine-grained texture cues and deep semantic information. Second, a C2CGA context-guided attention module is introduced to emphasize floral structural features while suppressing background noise caused by reflections and canopy clutter. Third, extended multi-scale detection heads (P2-P6) incorporate broader contextual information to reduce missed detections and false positives in dense planting scenarios. Experimental results on a custom chili flower dataset show that the proposed method achieves 92.8% precision, 83.7% recall, 90.8% mAP50, and 72.7% mAP50-95, improving upon the YOLOv11n baseline by 2.1, 1.3, 3.9, and 6.6 percentage points, respectively. Deployment on an NVIDIA Jetson AGX Orin edge platform demonstrates real-time inference at 97.9 FPS, confirming its suitability for on-device phenotyping. Overall, the proposed approach significantly improves detection robustness under complex greenhouse conditions, providing an effective visual sensing methodology for automated crop monitoring and data-driven yield estimation.

Why it matches plant phenotyping methodsチリの花・蕾という植物器官の画像検出手法を中心に、モデル改良、専用データセットでの性能評価、エッジ実装まで行っており、植物表現型取得の技術的貢献が明確である。

titleGCP-YOLO: high-precision detection of tiny chili flowers in complex greenhouse scenes.

PhenoNEST: A Neuro-Symbolic Framework for Ontology-Aware Multimodal Plant Phenotyping and Trait Discovery →

High-throughput plant phenotyping generates valuable data that often remains trapped in unstructured text and isolated RGB images. To bridge this semantic gap, we propose a framework for constructing a multimodal granular Knowledge Graph (KG) to monitor genotype-phenotype interactions across time and experiments. In this work, we focus on wheat Triticum aestivum as a representative target crop to validate our methodology across complex canopy environments. Our pipeline first distills noisy field notes to extract entities and relations, dynamically constructing the KG by converting unique instances into hierarchical class entities via RDF-typing. These graph nodes are then aligned with standardized ontologies (PO, RO, WTO) using PlantDeBERTa. To visually ground the constructed graph, a Vision-Language Model paired with a wheat-segmentation ViT generates attention-based softmaps, linking specific KG entities directly to image pixels. We introduce a central observation node Plant_Obs_Id to connect these multimodal subgraphs temporally. Evaluated on 500 curated WisWheat samples using Pointing Game accuracy, Visual Word Sense Disambiguation (VWSD), and rank-based metrics, our neuro-symbolic approach successfully maps complex field observations to a structured graph. This enables automated field note auditing, temporal stress monitoring, and precise spatial trait localization for wheat breeders.

Why it matches plant phenotyping methods植物のマルチモーダル表現型データを知識グラフと画像に統合し、画像画素への形質局在化を行う中核的な計算フレームワークを提案・評価しているため。

abstractwe propose a framework for constructing a multimodal granular Knowledge Graph (KG) to monitor genotype-phenotype interactions across time and experiments

The Turning Point of 3D Plant Phenotyping: 3D Foundation Models Enable Minute-to-Second Cross-Crop Reconstruction and Beyond →

3D plant phenotyping is notoriously known to be procedure-complicated and of low throughput due to the extensive multi-view imaging, the fragile 3D reconstruction pipeline, and the additional cost from reconstructed geometry to phenotypic extraction. These limitations are further amplified in low-cost data acquisition, where smartphone videos or sparsely sampled multi-view images provide limited view overlap and self-occlusion. In this work, we show that the conventional 3D plant phenotyping pipeline could be streamlined and significantly accelerated with 3D Foundation Models (3DFMs), and particularly, present one of the first cross-crop 3D phenotyping frameworks powered by 3DFMs. The framework replaces COLMAP-style sparse initialization with 3DFM-based feed-forward geometric recovery, combines geometry-constrained 3D Gaussian Splatting for dense reconstruction, enables few-view reconstruction through iterative view synthesis and refinement, and converts reconstructed geometry into measurable organs through 2D-to-3D semantic transfer, metric scale recovery, and organ instance separation. We further construct a cross-crop dataset with smartphone-based image acquisition, diverse plant morphologies, and manual annotations for segmentation and phenotypic evaluation. Experiments across 26 plant sequences show that 3D Foundation Models reduce the average reconstruction time from 6.52 minutes to 1.58 seconds while maintaining high reconstruction quality and phenotyping accuracy. These results suggest a fresh technical route for high-throughput 3D plant phenotyping, from low-cost image acquisition to fast reconstruction, perception, scale recovery, and phenotypic measurement.

Why it matches plant phenotyping methods3D画像再構成から器官分離・形質測定までを統合した高速植物フェノタイピング手法を開発し、複数作物・データセットで性能評価しているため。

abstractpresent one of the first cross-crop 3D phenotyping frameworks powered by 3DFMs

A multiregional image–text dataset and benchmark for vision-language modeling of plant diseases →

Plant diseases remain a major challenge to global food production, and timely, accurate, and scalable detection of plant stress is critical to reducing these losses. Recent advances in digital imaging and artificial intelligence offer unprecedented opportunities for precision crop disease detection and management. Yet, existing plant disease datasets remain often fragmented across crop and disease systems, and are largely dominated by controlled-environment imagery. The lack of standardized, interoperable, and representative datasets limits reproducibility, transferability, and scalability of AI systems, thereby constraining their deployment in operational agricultural applications. Here we present LeafMD, an integrated multimodal plant disease dataset and benchmark resource that includes LeafNet 2.0, a large-scale multimodal digital image dataset comprising 255,855 image–text pairs across 37 crop species, 197 crop–disease classes, and 9 geographic regions spanning tropical, subtropical, and temperate agricultural systems. Unlike conventional datasets, LeafNet 2.0 integrates biologically grounded symptom descriptions with image-level annotations of early and late disease stages, enabling symptom-aware analysis of disease progression under realistic field conditions. We further introduce LeafBench 2.0 as part of LeafMD, a visual-question answering benchmark covering nine fine-grained plant pathology tasks, including pathogen classification, lesion characterization, symptom interpretation, and disease severity assessment. Evaluation across 16 vision–language models revealed substantial performance gaps between coarse disease recognition and fine-grained pathological reasoning, while agriculture-adapted models consistently outperformed several larger general-domain architectures on symptom-oriented tasks. Together, LeafNet 2.0 and LeafBench 2.0 establish LeafMD as a multimodal resource for developing disease-aware agricultural foundation models and studying fine-grained pathological reasoning in real-world environments.

Why it matches plant phenotyping methods植物病害の画像・症状記述データセットとベンチマークを構築し、病徴解釈・病変特徴・病害重症度評価を対象にモデル性能を評価しており、植物状態の取得・評価手法が中心である。

abstractHere we present LeafMD, an integrated multimodal plant disease dataset and benchmark resource

Field Dataset Construction and Real-Time Object Detection for Strawberry Leaf Disease Monitoring in Drone-Based Precision Spraying →

Abstract has not been obtained from indexed metadata or an accessible article page.

Why it matches plant phenotyping methodsイチゴ葉の病害状態を画像から検出するデータセット構築とリアルタイム手法が題名上の中心であり、植物病害フェノタイピングおよび評価基盤に該当する。

titleField Dataset Construction and Real-Time Object Detection for Strawberry Leaf Disease Monitoring in Drone-Based Precision Spraying

Design and Development of a Deep Learning-Based System for Multi-Fruit Disease Classification and Severity Detection Using VGG-16 and VGG-19 Architectures on an Expert-Verified Indian Fruit Crop Dataset →

Fruit diseases caused by fungal, bacterial, and viral pathogens cause devastating pre- and post-harvest losses to Indian agriculture, particularly in Maharashtra where Guava, Mango, Orange, Papaya, and Pomegranate are major horticulture crops. Accurate, early-stage disease identification directly impacts farmer income, food security, and precision crop management. Conventional manual inspection by agronomists is subjective, time-consuming, and not scalable across thousands of orchard acres. Automated deep learning-based image analysis has emerged as a transformative solution, offering high accuracy, speed, and field deployability. Existing deep learning models for plant disease detection are predominantly trained on the real time database, which inadequately represents Indian fruit crop species. Furthermore, published systems focus on binary disease presence detection and lack disease severity grading — a critical requirement for treatment decision-making. The absence of expert-verified datasets and the visual overlap among disease classes (such as Phytophthora vs. Scab in Guava, and Powdery Mildew vs. Ring-spot in Papaya) present significant classification challenges. Limited training data for rare classes such as Stylerandroot (Guava, 310 samples) and Bacterial Blight (Pomegranate, 304 samples) further compounds model generalization. This study presents an end-to-end deep learning framework using VGG-16 and VGG-19 architectures trained on a novel, expert-verified dataset of 7,372 approved images spanning 22 disease/healthy classes across 5 fruit types, after on filed validation. The pipeline includes image preprocessing (CLAHE, Gaussian denoising, normalization), hybrid multi-feature extraction (CNN features, GLCM, LBP, Color Histograms), transfer learning with progressive fine-tuning, and a cascaded severity estimation module. Three algorithms are designed: (1) a Transfer Learning Classification Algorithm using VGG-16/VGG-19 backbone with softmax multi-class head, (2) a Hybrid Feature Fusion Algorithm combining CNN deep features with handcrafted descriptors for improved minority-class performance, and (3) a Cascaded Rule-CNN Severity Estimation Algorithm classifying disease progression into Healthy, Mild, Moderate, and Severe categories. VGG-19 achieved 96.1% overall accuracy, 95.1% precision, 94.5% recall, and a macro F1-score of 0.942, significantly outperforming VGG-16 (94.5% accuracy, F1: 0.918). Mango classification achieved the highest accuracy at 97.8%, while severity estimation reached 91.2% overall accuracy with the Mild category being the most challenging at 86.4%.The proposed system demonstrates that expert-verified, domain-specific datasets combined with transfer learning and hybrid feature fusion significantly advance the state of fruit disease detection for Indian agriculture. This framework provides a scalable, interpretable, and practically deployable solution for precision horticulture.

Why it matches plant phenotyping methods植物画像から病害状態と重症度を推定する深層学習手法、データセット、評価を中心的に開発しており、植物表現型計測の方法論的研究に該当する。

abstractThis study presents an end-to-end deep learning framework using VGG-16 and VGG-19 architectures trained on a novel, expert-verified dataset of 7,372 approved images spanning 22 disease/healthy classes across 5 fruit types

Detection and Classification of Plant Leaf Diseases →

Plant diseases pose a serious threat to global food production, leading to yield losses, increased production costs, and environmental damage. Plant pests and diseases have widespread negative impacts on economic, ecological, environmental, and human health, and calculating the true cost of these damages is extremely difficult. Traditional diagnostic methods also present significant limitations in terms of time and cost. This study adopts an image processing-based approach to this problem. Using the Crop Disease Detection dataset, plant disease detection was performed using CNN, DNN, K-NN, SVM, XGBoost, and Random Forest algorithms, employing both deep learning and machine learning methods. The study demonstrates that CNN architectures designed from scratch, without resorting to pre-trained models such as ResNet and MobileNet, can also exhibit high performance. The highest accuracy rate was obtained with the CNN model at 94.08%. In machine learning models, grid search was used for hyperparameter optimization, and the best results were achieved through this method.

Why it matches plant phenotyping methods植物葉の画像から病害状態を推定する画像処理・機械学習手法を比較評価しており、病害フェノタイピング手法が研究の中心である。

abstractThis study adopts an image processing-based approach to this problem.

SLIF-Tomato: Inverted Residual Convolutional Block Attention Module for In-field Tomato Leaf Disease Recognition →

Abstract Tomato cultivation represents a critical component of nutrition, economic development, and public health, yet it is increasingly compromised by foliar diseases that diminish yield and intensify dependency on hazardous agrochemicals. Although deep learning models have demonstrated strong capabilities for automated disease recognition, existing benchmark datasets exhibit limited real-world utility, primarily due to the absence of in-field imagery and precise annotations of diseased regions. A novel attention mechanism, I nverted R esidual C onvolutional B lock A ttention M odule ( IR-CBAM ), is proposed, combining inverted residual blocks with the CBAM Module, and is specifically tailored to address challenges posed by in-field image variability, such as complex backgrounds and inconsistent lighting. Furthermore, this study introduces SLIF-Tomato , the S ri L ankan I n- F ield Tomato leaf disease dataset, which is the first complete in-field dataset comprising class labels and bounding box annotations collected under diverse real-world conditions. The proposed approach achieved 99.66% and 99.91% accuracy rates on two curated versions of the SLIF-Tomato dataset. Subsequently, the YOLOv12-large model is employed to detect diseased regions, which obtained an average precision score of 88.5%. These contributions advance the development of accurate, efficient and field-adaptable diagnostic systems for tomato leaf disease management in precision agriculture.

Why it matches plant phenotyping methodsトマト葉の病害領域を画像から検出・認識する手法を開発し、実圃場データセットも構築・評価しており、植物病害状態の表現型取得が中心的です。

abstractA novel attention mechanism, I nverted R esidual C onvolutional B lock A ttention M odule ( IR-CBAM ), is proposed, combining inverted residual blocks with the CBAM Module, and is specifically tailored to address challenges posed by in-field image variability, such as complex backgrounds and inconsistent lighting.

DFA-YOLO: an enhanced YOLOv11-OBB and knowledge distillation-based maize stomata detection system. →

Introduction Stomata are vital gatekeepers of plants that regulate the fundamental trade-off between carbon gain and water loss. Precise, high-throughput identification of stomatal traits is therefore essential for assessing plant stress tolerance and water-use efficiency. However, conventional bounding box detection struggles to accurately localize densely distributed and arbitrarily oriented stomata. Methods This study proposes DFA-YOLO, an enhanced YOLOv11-OBB (Oriented Bounding Box) model for orientation-aware maize stomatal localization and preliminary OBB-derived trait extraction. Based on a maize stomatal dataset expanded from 1,053 to 3,597 microscopic images, DFA-YOLO integrates three task-specific components: (1) a cross-dataset MGD distillation strategy that transfers structural priors from single-stomata images to dense multi-stomata scenes; (2) a fixed-threshold Focaler-CIoU localization-loss reweighting strategy (u stomata = 0.95, d stomata = 0.00) to emphasize hard positive samples during OBB regression; and (3) a C3k2_AssemFormer feature-aggregation module that combines convolutional local feature extraction with linear attention-based context aggregation. Results DFA-YOLO achieved 94.1% mAP50, 84.8% mAP75, 74.1% mAP50-95, and 90.0% recall, with higher recall, mAP50, mAP75, and mAP50-95 than the YOLOv11-OBB baseline under the same OBB evaluation protocol. When deployed on an automated platform, the system processed images at 44.9 FPS and supported stomatal localization, density estimation, and preliminary orientation-aware size description. Discussion Under the tested maize microscopic imaging workflow, DFA-YOLO enables rapid extraction of detection-oriented stomatal traits and provides a prototype tool for high-throughput maize stomatal phenotyping.

Why it matches plant phenotyping methodsトウモロコシ気孔の画像検出・配向局在化と形質抽出を目的とするYOLOベース手法を開発し、データセット、精度比較、自動化プラットフォームでの性能を評価しているため、植物フェノタイピング手法が中心である。

abstractThis study proposes DFA-YOLO, an enhanced YOLOv11-OBB (Oriented Bounding Box) model for orientation-aware maize stomatal localization and preliminary OBB-derived trait extraction.

SeedMatExplorer: the transcriptome atlas of Arabidopsis seed maturation. →

Background Seed maturation is a critical developmental phase during which seeds acquire traits essential for nutritional value, desiccation tolerance, and long-term survival. Abscisic acid (ABA) signalling is a key regulator of this process, coordinating gene expression programs underlying the acquisition of seed quality traits. However, the molecular regulation of many of these traits remains poorly understood. To address this, we performed a comprehensive analysis of seed maturation in Arabidopsis thaliana, combining physiological and transcriptomic approaches across wild-type plants and mutants affected in ABA biosynthesis, signalling, and catabolism. Results We generated a high-resolution transcriptome dataset covering seed development from 12 days after pollination to the dry seed stage in wild-type and ten mutant lines. In parallel, we characterized the temporal acquisition of multiple seed traits, including germination capacity, dormancy, chlorophyll fluorescence, longevity and desiccation tolerance. Integration of these datasets using weighted gene co-expression network analysis (WGCNA) identified gene modules associated with specific trait acquisition patterns. This approach enabled the identification of coordinated transcriptional programs linked to distinct seed quality traits, extending beyond individual gene-level analyses. Notably, modules associated with desiccation tolerance and longevity were enriched for genes involved in stress responses and ABA-regulated pathways, highlighting the complex and multifactorial regulation of these traits. Conclusions This study provides a comprehensive physiological and transcriptomic framework for understanding seed maturation and the acquisition of key seed quality traits in Arabidopsis thaliana. By linking gene expression dynamics to trait development, our work offers new insights into the regulatory networks underlying seed resilience and storage capacity. The dataset is made accessible through SeedMatExplorer (https://www.bioinformatics.nl/SeedMatExplorer), an open-access web platform that enables interactive exploration and supports hypothesis generation. Together, this resource represents a valuable tool for advancing research on seed biology and improving seed performance in agricultural contexts.

Why it matches plant phenotyping methods種子成熟に伴う複数の植物形質を体系的に取得し、トランスクリプトームと統合した再利用可能なデータセットおよび探索プラットフォームを提供しており、単なる生物学的実験の routine 測定を超える。

abstractIn parallel, we characterized the temporal acquisition of multiple seed traits, including germination capacity, dormancy, chlorophyll fluorescence, longevity and desiccation tolerance.

DESENVOLVIMENTO DE APLICATIVO COM APRENDIZADO DE MÁQUINA PARA PREDIÇÃO DO IMPACTO DE RIZOBACTÉRIAS NO CRESCIMENTO DE PLANTAÇÕES DE ARROZ →

O uso de rizobactérias promotoras de crescimento de plantas (RPCPs) apresenta-se como alternativa sustentável para a agricultura, porém a predição de seus efeitos envolve múltiplas variáveis. Este trabalho teve como objetivo desenvolver um aplicativo móvel apoiado por aprendizado de máquina para análise preditiva do impacto de RPCPs no crescimento de arroz. A metodologia abrangeu quatro fases: levantamento de requisitos com especialista, análise exploratória de uma base de dados com 6.038 registros experimentais, desenvolvimento e avaliação de modelos de classificação e implementação do sistema. Foram comparados os algoritmos KNN, Random Forest e XGBoost, sendo este último selecionado por apresentar maior acurácia (0,945) e menor desvio padrão (0,010) na validação cruzada. A arquitetura Cliente-Servidor integrou um aplicativo Android em Kotlin com Jetpack Compose a uma API RESTful em FastAPI, operando em duas modalidades: não destrutiva, baseada em medições de campo, e destrutiva, com métricas de biomassa seca. Os resultados indicam que a ferramenta pode auxiliar a tomada de decisão ao reduzir a necessidade de coletas destrutivas em determinadas situações, contribuindo para práticas agrícolas mais sustentáveis.

Why it matches plant phenotyping methodsイネの生育影響という植物形質を、非破壊測定および乾物バイオマスから機械学習で予測するアプリケーションの開発・評価が研究の中心であり、単なる生育実験ではない。

abstractdesenvolver um aplicativo móvel apoiado por aprendizado de máquina para análise preditiva do impacto de RPCPs no crescimento de arroz

Cerrado plant traits (CPT): a database of functional traits across vegetation types in a global biodiversity hotspot →

Abstract Background Trait-based ecology has become central for understanding plant form, function and ecosystem processes, but progress has been hampered by biased representation in trait databases. As such, global trait syntheses remain strongly biased towards temperate forest biomes. Tropical savannas are the most extensive, biodiverse and disturbance-driven ecosystems worldwide, yet are poorly represented in functional trait databases, limiting ecological inference and applied decision-making. Scope Here, we introduce the Cerrado Plant Traits (CPT), an open-access initiative compiling and standardising plant functional trait data for the Brazilian Cerrado, the world’s most biodiverse tropical savanna. CPT integrates trait information for all major plant organs (whole-plant, root, shoot, leaf, flower, fruit and seed) across vegetation types in the Cerrado, drawing on a collaborative and inclusive research network. The current version of CPT compiles data from 148 datasets, totalling 113,859 curated trait records for 2,134 taxonomically verified species across 150 families. Trait records span pristine, degraded and restored environments and capture both interspecific and intraspecific variation. Whole-plant and leaf traits dominate the current dataset, while belowground and reproductive traits remain comparatively underrepresented, highlighting key priorities for future research. Conclusions By substantially increasing the representation of savanna species in global trait repositories, CPT enables tests of ecological hypotheses across multiple levels of organization, analyses of trait–environment relationships across fire, soil and climatic gradients, and robust comparisons across forest–savanna transitions. Beyond its scientific value, CPT provides a practical, standardised resource to support conservation planning, restoration programs and evidence-based policy in a biodiversity hotspot facing accelerating land-use and climate pressures.

Why it matches plant phenotyping methods植物の機能形質を標準化して統合した大規模データセットであり、再利用可能な形質リソースの構築が中心です。

abstractHere, we introduce the Cerrado Plant Traits (CPT), an open-access initiative compiling and standardising plant functional trait data for the Brazilian Cerrado

Challenges in Maize Root Phenotyping: Preprocessing Limits and Class Imbalance in Deep Learning →

Doubled Haploid (DH) technology significantly accelerates the development of homozygous lines in maize breeding; however, its scalability is constrained by the reliable discrimination of haploid and diploid individuals. The widely used R1-nj anthocyanin marker at the seed stage is susceptible to genetic suppression and environmental variability, leading to high misclassification rates. This limitation has driven a shift toward seedling root morphology as a more robust phenotypic marker, yet it introduces major challenges, including complex image noise and severe class imbalance. In this study, we systematically evaluate the limitations of standard computer vision pipelines and baseline deep learning models for root-based classification. Automated background removal methods (HSV, Rembg) are shown to misinterpret fine root hairs as noise, resulting in significant morphological data loss. Additionally, experiments conducted under a realistic class imbalance (1:5.4) demonstrate that widely used CNN architectures (ResNet50, VGG16, EfficientNetB0, DenseNet121) exhibit strong majority class bias, with haploid recall dropping to 0.00% and 27.7%. These findings reveal a critical limitation in existing approaches and highlight the need for domain-informed datasets and imbalance-aware learning strategies for robust and scalable AI-based maize breeding systems.

Why it matches plant phenotyping methodsトウモロコシの根形態画像を用いた倍加半数体・二倍体分類について、画像前処理と深層学習モデルの限界を体系的に評価しており、表現型取得・抽出手法が中心である。

abstractIn this study, we systematically evaluate the limitations of standard computer vision pipelines and baseline deep learning models for root-based classification.

AnnonFruitTraits 1.0, a comprehensive dataset on frugivory-related traits for Annonaceae species worldwide →

Abstract Functional traits are critical for understanding species interactions within ecosystems and their responses to environmental changes. Yet, traits related to fruits and seeds are still underrepresented, especially in tropical ecosystems where mutualisms between fruits and fruit-eating animals are prominent. Here, we introduce AnnonFruitTraits 1.0, a comprehensive dataset of 34,772 records encompassing 26 frugivory-related traits for 2,266 species (ca. 90% of total species) of the pantropical plant family Annonaceae (Magnoliales). This dataset includes trait definitions and their significance for frugivory, as well as a description of our workflow from data acquisition to visualization. To facilitate data accessibility and reproducibility, we provide an accompanying R package (AnnonTraits) that enables users to explore, summarise, and visualise the dataset. By assessing species and trait coverage across genera and regions, we identified major data gaps in the Asia-Pacific region and in several Annonaceae genera (e.g., Artabotrys , Miliusa , Orophea, Polyalthia , and Uvaria ). Our findings show the importance of expanding trait data collection and taxonomic efforts, particularly in underrepresented regions and lineages. AnnonFruitTraits is a valuable resource for advancing research on seed dispersal, plant–animal interactions, and tropical forest conservation.

Why it matches plant phenotyping methods植物の果実・種子形質を大規模に整理した再利用可能なデータセットであり、データ取得から可視化までのワークフローと探索・要約・可視化用Rパッケージを提供しているため、形質データ資源として中心的です。

abstractHere, we introduce AnnonFruitTraits 1.0, a comprehensive dataset of 34,772 records encompassing 26 frugivory-related traits for 2,266 species

Benchmarking hybrid CNN and transformer backbones with graph convolution networks (GCN) for flower growth-stage classification. →

Accurate recognition of flower growth stages is important for plant phenotyping but remains challenging due to subtle visual differences and limited labeled data. This study proposes a hybrid CNN/Transformer + GCN framework for fine-grained flower growth-stage classification. A new dataset, BD Flower Growth, is introduced with 3,889 original images from eight Bangladeshi flower species, categorized into three stages (early, mid, full), forming 24 classes. The dataset is divided into training and testing sets, with augmentation applied only to the training data. Deep backbone networks are used to extract feature maps, which are transformed into graph representations and refined using Graph Convolutional Networks (GCN). A systematic ablation study is conducted by varying GCN depth (3, 5 layers), node resolution ([Formula: see text], [Formula: see text]), and graph construction methods (4-neighbour, 8-neighbour, and KNN with [Formula: see text]). Experimental results show that performance depends strongly on both backbone and graph configuration. The best performance of 97% accuracy is achieved by EfficientNetV2, DenseNet201-based hybrid models, additionally Swin Transformer model shows the largest improvement, increasing from 84% to 97% after GCN integration. Across different settings, grid-based graphs (4- and 8-neighbour) consistently provide more stable and higher performance compared to KNN graphs, while moderate GCN depth (3-5 layers) offers the best balance accuracy. Cross-dataset evaluation on the Oxford 102 Flower dataset further demonstrates the generalization capability of the proposed approach. These findings highlight the effectiveness of hybrid graph-based learning and the importance of graph configuration in improving fine-grained classification.

Why it matches plant phenotyping methods花の生育段階という植物状態を画像から分類する手法の開発・比較検証が中心で、新規データセットとアブレーションおよびクロスデータセット評価も含むため。

abstractAccurate recognition of flower growth stages is important for plant phenotyping

OPTIFARM: Benchmarking YOLO Architectures for Location-Robust Potato Quality Detection. →

Potato sorting in post-harvest processing relies heavily on manual visual inspection, which is physically demanding, subjective, and insufficiently scalable for modern packing lines. This study investigates the feasibility of a low-cost RGB-based optical inspection system for automated potato quality detection using deep learning-based object detection. A controlled imaging platform was constructed using commodity hardware, and a dataset of 19,805 manually annotated instances across 1361 images was collected from two geographically distinct farm locations in Slovenia. A systematic benchmark of 25 model configurations spanning five YOLO architecture families-YOLOv8, YOLOv9, YOLOv10, YOLOv11, and YOLO26-was conducted across three practical quality classes (Edible, Feed, Rotten) using a strict cross-location evaluation protocol in which models were trained on one location and tested on a completely unseen second location. All models achieved strong in-distribution performance (F1 ≥ 0.906), but showed considerable variation under cross-location conditions, with external F1 ranging from 0.792 to 0.918. The yolo26_l configuration achieved the best cross-location performance (F1 = 0.918, mAP@0.5:0.95 = 0.816, ΔF1 = 0.029), demonstrating that transferable representations are achievable under a standard supervised training protocol. Per-class analysis identified feed detection as the primary generalization bottleneck. The results confirm that affordable RGB-based sorting systems are technically feasible and highlight cross-location evaluation as an essential protocol for assessing real-world deployment readiness.

Why it matches plant phenotyping methodsジャガイモ塊茎の品質・腐敗状態をRGB画像と物体検出で推定する撮像システムを構築し、複数YOLOモデルの交差地域ベンチマークと外部検証を行っており、表現型取得法が中心である。

abstractThis study investigates the feasibility of a low-cost RGB-based optical inspection system for automated potato quality detection using deep learning-based object detection.

Software and reproducibility package →

First stable release of the RGB and NPQ pixel-wise phenotyping pipeline associated with the manuscript "Image-based biomarkers effectively predict salt and drought stress in dwarf tomatoes (Solanum lycopersicum L.)". This repository includes a Python-based image analysis pipeline for high-throughput plant phenotyping using RGB and chlorophyll fluorescence (NPQ) imaging data. The workflow is designed for pixel-wise extraction and analysis of image-derived traits, with a specific focus on preserving full spatial distributions rather than relying on image-level summary statistics. The pipeline processes RGB images to compute vegetation indices derived from color channel combinations, and NPQ fluorescence images to extract pixel-level chlorophyll fluorescence metrics. Both data types are integrated with experimental metadata through structured indexing files. The analysis framework is organised into two main stages: (i) data pre-processing and structuring into long-format pixel-wise datasets, and (ii) distribution-based statistical analysis of trait variability across treatments and conditions. The latter includes normalised histograms, Jensen–Shannon and Wasserstein distance metrics, and cluster-based permutation testing to identify statistically significant differences between distributions. A minimal example dataset is provided to enable end-to-end testing of the workflow, including image processing, metadata integration, and statistical analysis. An additional archive containing representative example outputs generated from the example dataset is included to illustrate the structure and format of intermediate and final pipeline outputs. To facilitate computational reproducibility, the repository also includes complete derived outputs generated from the full study dataset, including distribution-comparison results (Jensen–Shannon and Wasserstein distances) and cluster analysis outputs for all evaluated RGB and chlorophyll fluorescence traits. These files are provided as supplementary computational products of the workflow and can be used to verify, inspect, and reproduce the analyses described in the associated manuscript.

Why it matches plant phenotyping methodsRGBおよびNPQ画像から植物形質を画素単位で抽出・解析する再利用可能なパイプラインと再現性用データを提供しており、フェノタイピング手法が中心である。

abstractFirst stable release of the RGB and NPQ pixel-wise phenotyping pipeline

Automatic measurement of rice tiller angle from unmanned aerial vehicle images →

Abstract Rice ( Oryza sativa L.) tiller angle is an important trait that influences plant architecture, canopy light interception, and yield potential. In this study, we proposed a deep learning‐based pipeline for automated measurement of rice tiller angle and plant base width using unmanned aerial vehicle (UAV) imagery. Our method leverages keypoint detection models to estimate tiller angles efficiently and accurately. We collected and annotated a dataset of UAV‐captured rice plant images for tiller angle estimation. We demonstrated that our approach provides scalable and precise measurement for keypoints under real‐world field conditions, achieving a mean average precision (mAP)@50 of 0.982 and a mAP@50:95 of 0.859 on the test set. Predicted tiller angle distribution aligns well with human annotations, with a mean absolute error of 5.3° across a range of 10.7°–27.8° and a Pearson's correlation coefficient of 0.64, offering acceptable accuracy for tiller angle measurement in real‐world agricultural settings. Additionally, the predicted plant base width ranges from 2.8 to 5.5 cm, with a mean absolute error of 1.02 cm compared to human annotations, highlighting the model's capability for precise spatial analysis. Significant differences in tiller angle and plant base width were detected among 27 rice genotypes. These results validate the proposed pipeline's potential for accurate and efficient differentiation of plant architecture traits. This research is the first to measure the rice tiller angle directly from UAV images. It lays a foundation for automated phenotyping of plant architecture traits and has the potential for integration into plant phenotyping frameworks to further promote artificial intelligence‐driven rice research and production.

Why it matches plant phenotyping methodsUAV画像と深層学習により、イネの分げつ角度・株元幅という植物形態形質を自動推定する手法を開発・検証しており、フェノタイピング手法が研究の中心である。

abstractwe proposed a deep learning‐based pipeline for automated measurement of rice tiller angle and plant base width using unmanned aerial vehicle (UAV) imagery.

Software and reproducibility package →

First stable release of the RGB and NPQ pixel-wise phenotyping pipeline associated with the manuscript "Image-based biomarkers effectively predict salt and drought stress in dwarf tomatoes (Solanum lycopersicum L.)". This repository includes a Python-based image analysis pipeline for high-throughput plant phenotyping using RGB and chlorophyll fluorescence (NPQ) imaging data. The workflow is designed for pixel-wise extraction and analysis of image-derived traits, with a specific focus on preserving full spatial distributions rather than relying on image-level summary statistics. The pipeline processes RGB images to compute vegetation indices derived from color channel combinations, and NPQ fluorescence images to extract pixel-level chlorophyll fluorescence metrics. Both data types are integrated with experimental metadata through structured indexing files. The analysis framework is organised into two main stages: (i) data pre-processing and structuring into long-format pixel-wise datasets, and (ii) distribution-based statistical analysis of trait variability across treatments and conditions. The latter includes normalised histograms, Jensen–Shannon and Wasserstein distance metrics, and cluster-based permutation testing to identify statistically significant differences between distributions. A minimal example dataset is provided to enable end-to-end testing of the workflow, including image processing, metadata integration, and statistical analysis. An additional archive containing representative example outputs generated from the example dataset is included to illustrate the structure and format of intermediate and final pipeline outputs. To facilitate computational reproducibility, the repository also includes complete derived outputs generated from the full study dataset, including distribution-comparison results (Jensen–Shannon and Wasserstein distances) and cluster analysis outputs for all evaluated RGB and chlorophyll fluorescence traits. These files are provided as supplementary computational products of the workflow and can be used to verify, inspect, and reproduce the analyses described in the associated manuscript.

Why it matches plant phenotyping methodsRGBおよびNPQ画像から植物形質を画素単位で抽出・解析する再利用可能なパイプラインと再現性資料が中心であり、植物フェノタイピング手法に該当する。

abstractPython-based image analysis pipeline for high-throughput plant phenotyping using RGB and chlorophyll fluorescence (NPQ) imaging data.

A Weighted Ensemble of Convolutional Neural Networks for Anthracnose Detection in Avocado Fruit →

Global avocado production exceeds 10 million tons annually. Among the diseases affecting avocado fruit, anthracnose is one of the most significant, causing black lesions and fruit decay that can result in yield losses of 20–30%. To facilitate the early detection of anthracnose, this study proposes a computer vision-based approach. A dataset containing 2218 images of Fuerte avocados was first developed, comprising 1730 healthy samples and 488 anthracnose-infected samples after the labeling process. In the experimental phase, several convolutional neural network (CNN) models with varying depths (3, 4, 5, and 6 layers) were designed and evaluated. These models were subsequently integrated into different weighted ensemble configurations, where the best performance was achieved by the ensemble combining all four individual CNNs. The proposed weighted ensemble was compared against widely used state-of-the-art architectures, including VGG-16, ResNet-18, and MobileNetV2. Experimental results demonstrated the effectiveness of the proposed approach, achieving an F1-score of 0.9052, outperforming VGG-16 (0.8283), ResNet-18 (0.7328), and MobileNetV2 (0.7320).

Why it matches plant phenotyping methodsアボカド果実の病徴（炭疽病病変）を画像から検出するコンピュータビジョン手法を開発・比較検証しており、植物状態の取得が研究の中心である。

abstractTo facilitate the early detection of anthracnose, this study proposes a computer vision-based approach.

Automated phenotyping of soybean stomatal responses to water deficit using YOLOv8 →

Artificial intelligence applied to plant phenotyping is crucial for consistent results, as stomata classification under stress impacts physiology, water use efficiency, and productivity. Manual analysis is laborious and error-prone, limiting the efficiency and accuracy of evaluations. In this context, this study developed a soybean-specific dataset from water deficit (WD) and well-watered (WW) plants, training YOLOv8 model for automated detection and classification of open vs. closed stomata. Soybean plants were grown under water deficit and well-watered conditions, generating significant variations in stomatal structural opening and associated gas exchange traits. To capture stomata variations, epidermal printing techniques were employed, with images obtained by microscopy. The dataset was labeled using the intelligent polygon tool of the Roboflow application, with 269 images of the adaxial and abaxial surfaces of leaves annotated in two categories: open and closed stomata. The images underwent geometric transformations to facilitate model training. The results demonstrated that the YOLOV8 neural network achieved precision recall and mAP greater than 90%, highlighting its effectiveness in detecting and classifying stomata. By integrating automated classification of aperture states (open and closed) with a defined physiological stress context in soybean, this work establishes a dataset specifically designed for functional analysis. This approach extends the applicability of deep learning toward stress-oriented plant physiology studies, offering a robust tool for evaluating crop adaptation under climate change scenarios.

Why it matches plant phenotyping methodsヨロウ豆の気孔開閉状態を画像から自動検出・分類するYOLOv8手法とデータセットを開発・評価しており、植物表現型取得が研究の中心である。

abstractthis study developed a soybean-specific dataset from water deficit (WD) and well-watered (WW) plants, training YOLOv8 model for automated detection and classification of open vs. closed stomata.

Aerial imagery and deep learning accurately estimate maize foliar disease severity →

ABSTRACT Southern leaf blight (SLB) is a foliar disease of maize (Zea mays L.) caused by the necrotrophic fungal pathogen Cochliobolus heterostrophus. Genetic resistance is the most effective control method for SLB. Developing disease resistant maize lines requires field trials during which disease phenotypes must be visually assessed. Remote sensing using drones is an emerging technology that can be leveraged for high-throughput phenotyping of disease severity that is otherwise labor-intensive and subjective. This project used a deep learning approach to estimate SLB disease severity of single-row maize plots from drone imagery. Over 26,000 plot-level images produced from flights conducted across three growing seasons were labeled with in-field visual scores taken contemporaneously by expert raters. Variation in environmental conditions contributed to a labeled image dataset that reflects the complexity of agronomic field experiments. We assessed the ability of nine deep learning models from three architectural families to estimate disease severity. The best-performing model, EVA-02-B, achieved strong cross year generalization (R 2 = 0.697). Error analysis found that performance was more strongly associated with seasonal disease progression and flight-score time offset than with image-level noise. UAV-based deep learning estimated SLB severity with comparable precision to expert raters. This study lays the groundwork for integrating automated phenotypes into genetic studies of disease resistance. PLAIN LANGUAGE SUMMARY Southern leaf blight (SLB) of maize is a disease that causes yield loss worldwide and developing resistant varieties offers the best hope for controlling the disease. Studying SLB resistance requires plant pathologists to visually score severity in the field, a labor-intensive method that requires expertise. To address these challenges, we asked whether SLB severity scoring could be automated using drone images and artificial intelligence (AI). We trained AI models using three years of image and score data then compared the results to visual scores taken by five plant pathologists. The best performing AI model showed a similar level of consistency to the experts and proved capable of scoring severity despite unpredictable and uncontrollable conditions that affect field imaging experiments such as weeds or shadows. These findings provide a validated method that improves the efficiency of maize disease research, a critical area of study for agricultural sustainability and productivity.

Why it matches plant phenotyping methodsドローン画像と深層学習により、トウモロコシの葉病害重症度という植物状態を推定し、複数年データで性能と汎化性を評価した手法研究である。

abstractRemote sensing using drones is an emerging technology that can be leveraged for high-throughput phenotyping of disease severity

Design and Application of an Intelligent Plant Disease and Pest Recognition System for Landscape Architecture Based on Deep Learning →

Manual inspection of landscape pathology and textile fiber defects suffers from inherent subjective bias and suboptimal throughput. To bypass these bottlenecks, we propose the Ghost-Convolution Enlightened Vision Transformer (GeT). We constructed a novel hybrid neural network architecture, the Ghost-Convolution Enlightened Vision Transformer (GeT), which synergistically integrates the lightweight local feature extraction proficiency of Convolutional Neural Networks (CNN) with the global semantic modeling capabilities of Vision Transformers (ViT). Utilizing a newly established standard dataset (GLDP15k) comprising 15, 000 heterogeneous field images, the model was subjected to rigorous hyperparameter optimization and ablation studies. Optimization on the GLDP15k dataset yielded a peak accuracy of 96.8% across 12 target classes, maintaining a Kappa-coefficient of 0.941. Constrained to 1.16 M parameters, the architecture executes at 5.5 ms per image (180 FPS) on edge hardware. A 6-month application at Yuexiu Park demonstrated a 3.2-fold improvement in detection efficiency and a 35% reduction in pesticide usage compared to manual inspections. This study not only elucidates the interpretability of hybrid attention mechanisms in phytopathology but also adapts these vision-based paradigms to the detection of microscopic anomalies in textile weaving patterns, providing a scalable and computationally efficient solution for both precision plant protection and industrial fabric defect inspection.

Why it matches plant phenotyping methods植物病害を画像から認識する深層学習モデルを開発し、専用データセットで検証・応用しており、植物の病害状態の取得方法が中心的な研究貢献である。

abstractUtilizing a newly established standard dataset (GLDP15k) comprising 15, 000 heterogeneous field images, the model was subjected to rigorous hyperparameter optimization and ablation studies.

Uncertainty-Aware 3D Plant Reconstruction from Sparse Video Frames Using Neural Radiance Fields →

Abstract Automated three-dimensional reconstruction of plant architecture under- pins high-throughput crop phenotyping, yet its deployment in practical field settings is constrained by two fundamental limitations of Neural Radiance Fields (NeRF): degraded geometry under sparse, unstructured image capture, and a complete absence of calibrated uncertainty estimates that would allow practitioners to distinguish reliable geometry from reconstruction artefacts. We present UA-PlantNeRF, a unified framework that resolves both limita- tions through three tightly coupled contributions. First, building on VISAR, our prior intelligent video frame-selection strategy (weights α1 = 0.4, α2 = 0.3, α3 = 0.3), the pipeline identifies maximally informative, non-redundant view- points from raw footage with as few as 15 frames. Second, a dual-head NeRF architecture augmented with Monte Carlo Dropout produces jointly decom- posed aleatoric and epistemic uncertainty alongside each reconstructed voxel, trained under a heteroscedastic negative log-likelihood objective. Third, split conformal prediction—with calibration performed on held-out plants to preserve exchangeability— yields provable, distribution-free per-ray cov- erage guarantees at any user-specified confidence level. Evaluated across three benchmarks (Pheno4D, ROSE-X, and our novel UA-Field greenhouse dataset) at sparsity levels N ∈ {15, 30, 50}, UA-PlantNeRF achieves PSNR 28.7±0.6 dB and SSIM 0.89±0.01 at N = 30, outperforming all sparse-view baselines (p

Why it matches plant phenotyping methods植物の3D構造・アーキテクチャを推定する画像ベースのNeRF手法を開発し、複数ベンチマークで評価しているため、植物表現型計測法が中心です。

abstractEvaluated across three benchmarks (Pheno4D, ROSE-X, and our novel UA-Field greenhouse dataset)

Robust multi-target multi-scale tomato leaf disease detection for precision agriculture applications. →

The tomato is one of the most important economic crops worldwide; frequent occurrences of foliar diseases can severely affect its quality and yield, resulting in substantial economic losses. However, state-of-the-art methods still struggle with multi-target, multi-scale disease detection in complex scenarios, lacking accuracy and speed for tomato leaf diagnosis. A novel improved YOLO v8s model is proposed in this study to achieve high-precision and fast identification of multi-target and multi-scale tomato leaf diseases. First, a multi-target, multi-scale image dataset encompassing seven typical tomato diseases was developed to effectively enhance the model's robustness under complex practical scenario by integrating multiple public datasets and employing diverse data augmentation techniques. Second, a transfer learning strategy was employed to transfer high-quality features from a pretrained model to the disease detection task, thereby improving convergence speed and generalization ability. Finally, the CBAM (Convolutional Block Attention Module) channel-spatial attention mechanism was introduced into the YOLO v8s network, enabling the model to adaptively focus on critical regions and significantly enhance feature extraction and target localization performance. Experimental results demonstrate that the improved YOLOv8s-CBAM model achieves superior performance in complex scenarios, with a precision of 96.9%, recall of 97.3%, F1 score of 97.0%, and mAP@0.5 of 99.1%, representing improvements of 2.5%, 2.0%, 2.2%, and 1.8%, respectively, over the original YOLO v8s model. Moreover, the model size was reduced to 24.8 MB, a decrease of 11.7 MB compared to the original, achieving an effective balance between accuracy and lightweight design. These results indicate that the proposed method exhibits enhanced feature extraction and localization stability in multi-target, multi-scale disease identification tasks, providing an effective technical solution for automated detection in complex agricultural disease scenarios.

Why it matches plant phenotyping methodsトマト葉の病徴を画像から検出・識別するYOLOv8s-CBAM手法を開発し、データセット構築と性能比較検証を行っており、植物病害状態の画像ベース表現型計測が中心である。

abstractA novel improved YOLO v8s model is proposed in this study to achieve high-precision and fast identification of multi-target and multi-scale tomato leaf diseases.

Multi-Scale and Global–Local Feature Enhanced Detection for Tobacco Plants in Complex Field Environments →

Accurate detection of tobacco plants in complex field environments is critical for precision agriculture, crop monitoring, and yield estimation. Traditional manual counting methods are time-consuming, labor-intensive, and susceptible to environmental and subjective factors. In this study, we propose an improved YOLO11-based framework for automated tobacco plant detection, specifically designed to address challenges such as scale variation, dense distribution, and background interference. The framework integrates four key modules: the Edge-Enhanced Feature Stem (EEFS) to strengthen low-level feature extraction, the Multi-Scale Kernel Interaction (MSKI) to capture multi-scale contextual information, the Adaptive Weighted Feature Fusion (AWFF) to optimize feature aggregation, and the Global–Local Synergistic Attention (GLSA) to enhance feature discrimination by jointly modeling local details and global context. A comprehensive UAV-based tobacco dataset was constructed, encompassing multiple lighting conditions, collection heights, and observation angles. Experimental results demonstrate that the proposed method significantly outperforms the YOLO11 baseline and achieves superior performance compared to mainstream YOLO variants. Ablation studies and heatmap visualizations confirm the effectiveness of each module. Furthermore, the model exhibits robust performance under multi-dimensional environmental perturbations, including varying illumination, scale, and camera angles. The proposed framework provides a practical and efficient solution for automated tobacco plant counting, offering potential applications in UAV-based precision agriculture and large-scale crop monitoring.

Why it matches plant phenotyping methodsUAV画像からタバコ個体を自動検出・計数する手法を開発し、専用データセット、比較実験、アブレーション、頑健性評価まで行っており、植物個体数の取得方法が中心です。

abstractwe propose an improved YOLO11-based framework for automated tobacco plant detection

A Measurement-Driven Digital Twin Architecture for Plant-Level Biomass Estimation and Growth Forecasting in Hydroponic Systems →

Alternatives to soil-based horticulture, such as hydroponics, have been developed to respond to food distribution concerns for dense urban centers. A new system was developed to track an individual lettuce plant's growth in a hydroponic environment, utilizing streams of measured information and available models to continuously update the growth trajectory estimates for a plant. These "digital twin" models were integrated into an operating hydroponic greenhouse, with custom horticultural and sensor hardware to grow and measure relevant information. To aid in updating model parameters, plant yield was continuously measured with a custom neural network, using RGB-D images of the plants as an input. The network, trained on a collected dataset of 1300 images, was able to estimate mass within 1.5 g of the ground-truth value. After integration into the custom system, digital twin growth projections could approximate future yield between one and four days in the future, maintaining around a 2 g forecasting error.

Why it matches plant phenotyping methodsRGB-D画像から個体レタスの収量・質量を推定するニューラルネットワークと、センサー統合型の成長追跡基盤が研究の中心であり、植物形質の取得・予測手法を実質的に開発・検証している。

abstractA new system was developed to track an individual lettuce plant's growth in a hydroponic environment, utilizing streams of measured information and available models to continuously update the growth trajectory estimates for a plant.

From Occlusion to 3D: Amodal Completion Enables Single-View Wheat Reconstruction →

Abstract Occlusion is a major factor limiting accurate three-dimensional (3D) wheat phenotyping. In natural growth conditions, overlapping spikes, leaves, and stems often make only partial target regions visible in single-view images, hindering complete and reliable 3D reconstruction. To address this problem, this study proposes an amodal completion-driven framework for single-view 3D reconstruction of occluded wheat. The framework first uses visible prompts to recover the complete appearance and structural cues of occluded targets, and then feeds the completed images into single-view 3D reconstruction models to generate complete 3D structures. To support model training and evaluation, we construct the MMWO (Multi-view Multi-instance Wheat Occlusion) dataset from MMW by synthesizing diverse occlusion samples through organ-level cutouts, random geometric transformations, and region-constrained pasting, with annotations including visible masks, occlusion masks, and complete target images. Six representative reconstruction methods, including Direct3D, Real3D, SF3D, Spar3D, TRELLIS.2, and Hunyuan3D, are systematically evaluated. Hunyuan3D achieves the best geometric performance, with the lowest mean CD-L 1 and CD-L 2 values of 0.1286 and 0.0536, and the highest mean F-score of 0.5668. SF3D achieves the best rendering quality in terms of PSNR, SSIM, and LPIPS. In addition, Pix2Gestalt completion reduces the estimation errors of spike length, width, and area from 9.31%, 10.89%, and 32.23% to 4.64%, 9.70%, and 9.45%, respectively. These results demonstrate that amodal completion can effectively alleviate occlusion-induced information loss and provide more complete structural priors for single-view 3D wheat reconstruction. This study offers a feasible solution for robust 3D wheat phenotyping under occlusion and provides a systematic reference for applying general-purpose 3D generative models to agricultural phenotyping.

Why it matches plant phenotyping methods遮蔽下小麦の単視点3D再構成とアモーダル補完を開発し、データセット構築、複数手法の系統評価、器官形質推定誤差の検証まで行っており、植物フェノタイピング手法が中心である。

abstractthis study proposes an amodal completion-driven framework for single-view 3D reconstruction of occluded wheat

Toward smart agriculture: a hybrid mamba-transformer vision framework for plant disease detection →

Plant disease detection under complex field conditions remains a critical challenge for precision agriculture due to varying illumination, scale variations, subtle lesion patterns, and inter-class visual ambiguity. This study proposes MAFusionNet, a disease-aware hybrid vision framework integrating Mamba and Transformer architectures, with components explicitly designed for plant disease-specific challenges. The MAFusion Mixer operates parallel CS-Mamba and self-attention branches to simultaneously capture sequential lesion boundary evolution and global diseasecontext spatial relationships. The CS-Mamba branch employs the SS2D-LS Block with twodimensional selective scanning and Local-Selective enhancement for linear-complexity longrange modeling while preserving 2D lesion morphology. The PConv operator uses asymmetric directional kernels forming cross-shaped receptive fields to capture anisotropic disease patterns such as vein-aligned blights and directional rust streaks. We constructed PD40, a large-scale dataset with 80,369 expert-verified annotated images across 40 disease categories spanning eight major crops, with inter-annotator agreement Cohen’s κ = 0.874. Extensive experiments demonstrate that MAFusionNet achieves 94.7% mAP 50 and 81.8% mAP 50:95 on PD40, surpassing 25 state-of-the-art baselines including recent hybrid Mamba-Transformer detectors (CropMamba, HybridMamba, Mamba-DETR), with comprehensive ablation studies validating each component’s non-redundant contribution. Edge deployment analysis on NVIDIA Jetson hardware demonstrates practical feasibility: the compressed MAFusionNet-T-Lite variant (8.7M parameters) achieves 89.3% mAP 50 at 18.4 FPS on Jetson Nano with 8.3W power consumption. The dataset and code are available at PD40-Dataset GitHub Repository.

Why it matches plant phenotyping methods植物病害の症状を画像から検出・分類する視覚モデルを開発し、注釈付き大規模データセットで検証しているため、植物状態の画像ベース表現型計測が中心である。

abstractWe constructed PD40, a large-scale dataset with 80,369 expert-verified annotated images across 40 disease categories spanning eight major crops

Fine classification of rice diseases under field conditions based on improved ConvNeXt network →

Abstract Rice disease identification is a critical technique for ensuring yield and quality in precision agriculture. However, complex field backgrounds, subtle lesion features, and similar symptomatic manifestations have led to low accuracy and poor robustness in traditional classification methods. To address these issues, this study proposes an improved ConvNeXt network model (Improve‑ConvNeXt) for the fine classification of rice diseases under field conditions. A high‑quality dataset containing six categories (healthy rice, rice blast, brown spot, bacterial leaf blight, bacterial leaf streak, and bacterial grain rot) was constructed from field images and public datasets, with a total of 5,663 samples. Using ConvNeXt‑Tiny as the backbone, the model integrates a Hybrid Attention Transformer (HAT) to enhance the perception of lesion regions and key channels, and introduces Spatial and Channel Reconstruction Convolution (SCConv) to reduce feature redundancy and strengthen effective information expression. Experiments show that the Improve‑ConvNeXt model achieves 96.27% accuracy on the test set, which is 4.85% higher than that of the original ConvNeXt and significantly outperforms ResNet and DenseNet. The precision, recall, and F1‑score reach 95.84%, 96.11%, and 95.95%, respectively. Confusion matrix and Grad‑CAM visualization prove that the model can accurately focus on lesion areas and effectively distinguish similar diseases. This method provides high precision and strong generalization for rice disease identification in complex field environments, and offers a reliable technical reference for intelligent monitoring and precise management of rice fields.

Why it matches plant phenotyping methodsイネの病徴画像から病害状態を分類する画像・深層学習手法を開発し、データセットと性能比較で技術的に検証しているため、植物フェノタイピング手法が中心です。

abstractthis study proposes an improved ConvNeXt network model (Improve‑ConvNeXt) for the fine classification of rice diseases under field conditions.

Multi-Context Validation of Global Fractional Vegetation Cover Products in Croplands Using Multi-Source Crop FVC References →

Fractional Vegetation Cover of Crops (CropFVC) is a critical canopy parameter for monitoring crop growth, yet the behavior of widely used global FVC products (GLASS, GEOV1, GEOV2, and GEOV3) over croplands remains insufficiently understood due to fragmented validation references and limited crop-specific assessments. This study compiled a multi-source global CropFVC reference dataset (2000–2024) by integrating five international validation networks, the literature-derived samples, and newly acquired UAV and Jilin-1 satellite-derived CropFVC samples from China in 2024. The references were organized into three complementary validation contexts (V1~V3) to examine product behavior under different temporal coverage, crop purity, and reference conditions, together with spatio-temporal observations at the KONZ site. Results show that (1) across validation contexts, the evaluated products showed consistent behavior patterns, including shared overestimation under dense canopy conditions and reduced differences at low FVC levels; (2) spatio-temporal analysis at the KONZ site confirmed that peak-season deviations reflect shared response behavior rather than site-specific reference uncertainties; (3) historical mixed references (V1~V2) showed similar bias structures, whereas crop-specific validation (V3) preliminary revealed clearer crop-dependent responses, with predictive difficulty following winter wheat > maize > rice > soybean and improved stability after integrating 2024 observations. The integration of recent high-resolution crop observations expands existing global CropFVC references and enables behavior-oriented interpretation of global FVC products beyond simple accuracy ranking, providing an updated validation perspective for future development and application of global CropFVC products in agricultural monitoring.

Why it matches plant phenotyping methods作物キャノピーのFVCという植物形質を対象に、複数の全球FVC推定プロダクトを多様な参照データで体系的に検証し、UAV・衛星観測を含むCropFVC参照データセットを構築している。形質取得・検証が研究の中心である。

titleMulti-Context Validation of Global Fractional Vegetation Cover Products in Croplands Using Multi-Source Crop FVC References

YOLOv9-Based Detection of Diseases in Poplar Trees Using Histogram Equalization and Computer Vision. →

Poplar (Populus) trees are indispensable to various industries and environmental sustainability efforts. They are widely utilized for paper production, timber, and windbreaks, while also playing a significant role in carbon sequestration. Given their economic and ecological importance, the effective management of diseases is crucial. Convolutional Neural Networks (CNNs), renowned for their ability to process visual data, are pivotal in accurately detecting and classifying plant diseases. This study presents a domain-specific dataset of manually collected images of diseased poplar leaves from Uzbekistan and South Korea, ensuring geographic diversity and broader applicability. The dataset includes four disease classes, i.e., " Parsha (Scab) ," " Brown spotting ," " White-Gray spotting ," and " Rust ," which represent common afflictions in these regions. To advance research efforts, this dataset will be made publicly accessible, providing a valuable resource for the scientific community. Leveraging the cutting-edge YOLOv9c model, a state-of-the-art CNN architecture, we applied the Histogram Equalization technique as a preprocessing step to enhance the image quality to increase the accuracy of disease detection. This method not only improves the diagnostic performance of the model but also provides a scalable solution for monitoring and managing poplar diseases. By ensuring the health of poplar trees, this approach supports the sustainability of these critical resources. To our knowledge, this is the first publicly available dataset specifically focused on diseased poplar leaves, making it a significant contribution to global research efforts. It offers an invaluable resource for researchers and practitioners, enabling further advancements in early disease detection and sustainable forestry management.

Why it matches plant phenotyping methodsポプラ葉の病徴を画像から検出・分類する手法と公開データセットが研究の中心であり、植物の病害状態を直接推定するため、植物フェノタイピング手法として採択。

abstractThis study presents a domain-specific dataset of manually collected images of diseased poplar leaves from Uzbekistan and South Korea

LeafDet: A lightweight and interpretable deep learning framework for tomato leaf disease detection. →

Ensuring global food security depends on timely and reliable plant disease identification. Traditional disease detection methods often prove inefficient because of the lack of necessary precision. Furthermore, public datasets typically suffer from the class imbalance issue, which can obstruct reliable model testing and lead to biased performance evaluations. This paper introduces LeafDet, an object detection model based on the YOLOv8 architecture, specifically designed for the effective detection of tomato leaf diseases. Moreover, a revised, balanced dataset, named PlantTom, is developed by combining images from various public sources to reduce the existing dataset limitations. PlantTom has 7836 images with 8 distinct classes, each representing a tomato leaf disease. The proposed LeafDet model includes CBM, C2f, SPPF, and ECA attention modules in the backbone section; BiFPN, GSConv, VoVGSCSP, and Shuffle Attention in the neck section. Efficient attention methods like ECA and Shuffle Attention are used to improve both accuracy and speed. LeafDet model achieves 91.6% mAP@0.5 on the PlantTom dataset, which is a 2.2% improvement over the original YOLOv8n with 2.69M parameters and an inference time of 2.4ms. The proposed model also outperforms several other state-of-the-art object detection models, including the latest YOLOv11n and YOLOv12n. Ablation studies show that each part of the model helps to improve its performance, and the PIoUv2 loss function is found to be the optimal choice for this use. The model predictions are then validated using Eigen-CAM, which provides a visualization of the decision-making process. These results demonstrate that LeafDet provides a deployable and interpretable framework for plant disease detection in smart agriculture.

Why it matches plant phenotyping methodsトマト葉の病害状態を画像から検出する深層学習手法と評価用データセットを開発・検証しており、植物の表現型状態の取得が中心である。

abstractThis paper introduces LeafDet, an object detection model based on the YOLOv8 architecture, specifically designed for the effective detection of tomato leaf diseases.

Multisource Grapevine Phenology Dataset for Smart Farming and AI Modeling →

Abstract Artificial Intelligence and Machine Learning rely on large, high-quality datasets for accurate and robust models, yet data scarcity remains a major challenge, especially in smart farming. Phenology modeling, a key application, studies how plant biological events relate to climate and seasons. Accurate phenology models improve crop quality, support climate adaptation, and guide decisions such as pesticide use and harvesting, enhancing environmental and economic sustainability. However, agricultural data are highly diverse and heterogeneous, complicating model development. This study presents a proposed georeferenced dataset for Machine Learning-based grapevine phenology prediction across 3 Protected Designations of Origin in Aragón, Spain. Developed by a multidisciplinary team, the dataset combines 9 datasets from 8 sources (including meteorological time series, field phenology observations, and Copernicus Sentinel-2 multispectral imagery) covering the period 2016–2022. It supports both physical and Machine Learning-based phenology modeling and facilitates knowledge extraction in agronomy and plant biology. Its relevance lies in its comprehensive scope, the inclusion of 9 phenological stages, and a rigorous methodology ensuring reproducibility. This framework enables the creation of similar datasets for other regions or crops, advancing smart farming through scalable, data-driven solutions. The open publication of the code further supports this objective. We further anticipate its potential contribution to developing foundation models as well as to the creation of new knowledge in biology and agronomy.

Why it matches plant phenotyping methodsブドウの9段階のフェノロジーを対象に、圃場観測と衛星画像などを統合した再利用可能な地理参照データセットを構築しており、植物形質取得・モデル化の方法論が中心である。

abstractThis study presents a proposed georeferenced dataset for Machine Learning-based grapevine phenology prediction across 3 Protected Designations of Origin in Aragón, Spain.

LDTC-YOLO: A Lightweight Detection Model for Typical Citrus Leaf and Fruit Diseases in Real Orchard Environments →

Accurate detection of citrus leaf and fruit diseases is important for precision orchard management. However, real orchard images often contain small disease symptoms, leaf and fruit overlap, illumination variation, and cluttered backgrounds, making reliable detection challenging. This study proposes LDTC-YOLO, a lightweight YOLOv8n-based detection model for typical citrus leaf and fruit diseases in real orchard environments. To improve detection accuracy and model compactness, LDTC-YOLO integrates an Adaptive Feature Pyramid Network (AFPN) for cross-level feature fusion, Coordinate Attention (CA) for disease-region feature enhancement, a Lightweight Shared Convolutional Detection (LSCD) head for reducing parameter redundancy, and Wise-IoU (WIoU) for bounding-box regression optimization. In addition, a self-collected handheld citrus disease dataset, HOCD-4, was constructed using close-range smartphone images captured in real orchards. The dataset covers leaf and fruit symptoms of four typical citrus diseases: Huanglongbing/citrus greening (HLB), black spot, canker, and melanose. Experimental results show that LDTC-YOLO achieved precision, recall, mAP@0.5, and mAP@0.5:0.95 values of 0.915, 0.843, 0.894, and 0.648, respectively. Compared with YOLOv8n, LDTC-YOLO reduced parameters, GFLOPs, and model size from 3.006 M to 1.887 M, 8.1 to 7.4, and 5.97 MB to 3.83 MB, while increasing inference speed from 43.14 FPS to 47.45 FPS. These results indicate that LDTC-YOLO improves detection performance while maintaining a compact and efficient model profile, providing a potential reference for citrus disease detection under real orchard imaging conditions.

Why it matches plant phenotyping methods柑橘葉・果実の病徴を画像から検出する軽量モデルと実圃場データセットを開発・評価しており、植物の病害状態を推定するフェノタイピング手法が中心である。

abstractThis study proposes LDTC-YOLO, a lightweight YOLOv8n-based detection model for typical citrus leaf and fruit diseases in real orchard environments.

Multi-Scale Feature Rectification for Crop Leaf Disease Segmentation in Complex Scenarios →

Crop leaf disease segmentation in complex natural environments remains challenging because lesion regions often exhibit substantial scale variation, blurred boundaries, and severe background interference. To address these issues, this study proposes a Multi-Scale Feature Rectification Network (MFR-Net) for crop leaf disease segmentation. The proposed network adopts an EfficientNetV2-S-based encoder to extract hierarchical features, incorporates a hybrid attention mechanism to enhance lesion-sensitive spatial and channel representations, introduces a Cross-Window Atrous Spatial Pyramid Pooling (CWASPP) module to strengthen multi-scale contextual modeling, and employs a Feature Rectification Module (FRM) in the decoder to alleviate semantic inconsistency during cross-level feature fusion. Experiments on a Kaggle-derived benchmark constructed from the unaugmented data folder of the public Leaf Disease Segmentation Dataset, containing 588 diseased-leaf images and 588 corresponding binary lesion masks, showed that MFR-Net achieved the highest mIoU of 74.27% and the highest Recall of 87.61% among the compared methods, and maintained competitive Dice performance (84.25%) with 25.10 M parameters and 37.55 G FLOPs. Ablation results further confirmed the effectiveness of the proposed design, with CWASPP providing the most notable individual contribution. Additional experiments were conducted on an independent Apple Leaf Dataset comprising 3197 image–mask pairs, collected under mixed controlled and natural field-like imaging conditions. The results showed competitive performance under a different data distribution, and robustness evaluation further verified stable performance under severe noise, blur, darkness, and contrast variation. All experiments were implemented in PyTorch 2.11.0 (CUDA 12.8) on a workstation equipped with an NVIDIA GeForce RTX 4060 Ti GPU (8 GB). These results indicate that MFR-Net provides an effective and robust solution for crop leaf disease segmentation in complex agricultural scenarios.

Why it matches plant phenotyping methods病斑領域を画像から抽出するセグメンテーション手法を開発し、複数データセット、アブレーション、ノイズ等への頑健性で検証しており、植物病害状態の表現型取得が中心である。

abstractAblation results further confirmed the effectiveness of the proposed design

A Root Foundation Model for Zero-Shot Segmentation →

Foundation models pre-trained on massive datasets have demonstrated impressive performance, but in some specialised domains have been found to have lower accuracy. Domain-specific foundation models target a particular domain such as retinal or plant images. These domain-specific models have shown inconsistent results and the benefit to root segmentation is unknown. We train and evaluate the first domainspecific foundation model for root segmentation. Evaluation uses a leave-one-dataset-out design across nine diverse root datasets with two architectures. Applied zero-shot to unseen datasets, the root foundation model achieves 92% of fine-tuned Dice on average (0.636 versus 0.698), with 5 of 9 datasets above 90%. With 10 patches of few-shot fine-tuning, the root foundation model recovers 95% of its full-data Dice on average, versus 69% for a general pre-trained model. At low patch counts the general pre-trained model often failed to converge, with 5 of 9 datasets giving Dice below 0.05 at 3 patches, while the root foundation model produced Dice above 0.47 on every dataset and patch count. With full target-data fine-tuning, the two perform comparably, with mean improvements of +0.011 Dice for MobileSAM and +0.022 for M2F Swin-S, neither significant (Wilcoxon p = 0.150 and 0.064). We release our pre-trained MobileSAM root foundation model for use with RootPainter, enabling fully automatic root segmentation on new datasets with an ordinary laptop or desktop computer, with no need for annotation or training.

Why it matches plant phenotyping methods根の画像セグメンテーションを行う基盤モデルを開発し、9データセットでゼロショット性能を評価する研究であり、植物形質取得手法が中心です。

abstractWe train and evaluate the first domainspecific foundation model for root segmentation.

Tomato Ripeness Detection and Localization Based on the Intelligent Inspection Robot Platform. →

The field inspection and ripeness detection of tomatoes in China remain heavily dependent on manual labor, while existing robotic solutions often exhibit limited functionality, poor environmental adaptability, prohibitive hardware costs, and unstable positioning accuracy. To address these limitations, this study proposes an intelligent tomato inspection robot that seamlessly integrates real-time ripeness recognition with precise spatial localization. Built upon a Raspberry Pi 5 core controller, the robot employs a lightweight, layered modular architecture designed to flexibly navigate complex agricultural environments. A comprehensive, multi-dimensional image dataset of tomato ripeness was constructed to train a three-category detection model based on the YOLOv8n architecture. Following 413 training epochs, the model demonstrated exceptional performance, achieving an overall mAP@0.5 of 87.8% and an mAP@0.5:0.95 of 72.7% on the held-out test dataset. In field inspections, the system achieved detection precisions of 82.22% for immature tomatoes, 92.66% for half-ripened tomatoes, and 100% for fully ripe tomatoes, successfully identifying all ripe tomatoes and satisfying the practical demands of field inspection. Furthermore, the integration of an Ultra-Wideband positioning system yielded an overall Root Mean Square Error of 0.231 m, successfully confining positioning errors to within 0.24 m to fully satisfy the stringent localization demands of crop-level inspection. Field evaluations confirmed that under optimal configurations, the robot can efficiently inspect a 50-m planting row in 10 min (±1 min) and maintains a continuous operational battery life of 2 h (±10 min). The core contribution of this work is the system-level integration and optimization of technologies for greenhouse agriculture. This integrated design achieves low hardware cost and high deployment flexibility, addressing longstanding challenges of labor-intensive inspection and delayed harvesting, and delivering a practical solution for intelligent tomato plantation management.

Why it matches plant phenotyping methodsトマト果実の成熟状態を画像から推定するモデルと、データセット・ロボット検査プラットフォームを開発・評価しており、植物状態の取得方法が中心的です。

abstractThe core contribution of this work is the system-level integration and optimization of technologies for greenhouse agriculture.

In-Field Nondestructive Detection of Nitrogen Status on 'Yotsuboshi' Strawberry Using Deep Learning Algorithm. →

Nitrogen (N) management is critical for optimizing growth and fruit quality in open-field strawberry cultivation, demanding advanced technological solutions for reliable nutrient assessment. However, visual symptom diagnosis, though widely utilized for nutrient monitoring, is inherently subjective and prone to observer bias, resulting in inconsistent and often unreliable assessments. While available accurate tissue analysis is destructive and costly. Nondestructive, in-field imaging techniques such as the normalized difference vegetation index (NDVI) exist but require expensive multispectral imaging systems. To address these limitations, this study developed a streamlined methodology for in-field N status detection using deep learning on standard RGB images. The experiment utilized 'Yotsuboshi' strawberries in a randomized complete block design with sufficient nitrogen (T1) and deficient nitrogen (T2) treatments. To mitigate ambient light variability, a key challenge in open-field phenotyping, a low-cost phenotyping cylinder was developed for standardized smartphone image acquisition. Rigorous four-stage annotation criteria were also introduced to classify the nitrogen status in strawberry leaves as NormalN, LowN, or AdvancedLowN, ensuring a high-quality novel dataset. A YOLO11 model trained on this dataset achieved precision, recall, and mAP50 values exceeding 99%. Subsequent testing using the phenotyping cylinder yielded a mAP50 of 87%. In-field validation without a phenotyping cylinder also demonstrated robust performance under diffuse cloudy conditions (82.7% mAP50), outperforming direct sunlight (79% mAP50). Moreover, the model's classifications of 'NormalN' and 'LowN' statuses strongly corresponded with NDVI measurements, validating the accuracy of the RGB-based approach. This research demonstrates the significant potential of combining deep learning and phenotyping cylinder to create a rapid, low-cost, nondestructive and reliable tool for in-field nitrogen detection, with possible application across different crops and environmental conditions.

Why it matches plant phenotyping methods植物の窒素状態をRGB画像から推定する深層学習法、標準化撮像用デバイス、アノテーション基準、データセットを開発し、NDVIおよび圃場条件で検証しており、表現型取得・推定手法が中心である。

abstractthis study developed a streamlined methodology for in-field N status detection using deep learning on standard RGB images.

A Multi-Sensor UAS Platform: Design, Testing, and Application for High-Throughput Plant Phenotyping →

Dataset Structure and Concepts This dataset supports a study on the development and validation of a modular, regulation-compliant multi-sensor Unmanned Aerial System (UAS) for high-throughput plant phenotyping. The architecture utilizes a mid-lift UAV platform (Inspired Flight IF800) with a custom-engineered, decoupled payload system that separates sensing, power distribution, and onboard computing. The core concept focuses on an open-architecture design that ensures centimeter-level spatial co-registration across independent research-grade sensors. Contents of the Dataset Engineering Design Files: Full 3D assembly of the custom dovetail-mounted bracket and raised top platform in .STEP and SolidWorks (.SLDPRT) formats. Figures and tables in manuscript.

Why it matches plant phenotyping methods植物フェノタイピング用のマルチセンサーUASの設計・開発・検証が中心であり、センサー統合と空間共登録を含む再利用可能な測定プラットフォームを扱っている。

abstractthe development and validation of a modular, regulation-compliant multi-sensor Unmanned Aerial System (UAS) for high-throughput plant phenotyping

Identifying Nutrition Deficiency in Paddy Leaf using Neural Network →

Agriculture is the primary source of livelihood for majority of India’s population, with paddy serving as a staple food for a large segment of people. However, paddy cultivation is affected by several challenges that vary with climate, location, and farming practices. Among these, nutrient deficiencies in paddy leaves significantly impact crop yield and quality, making early detection crucial for effective farm management. The following study presents a novel approach to identifying nutrient deficiencies using neural networks and also provides a solution for their early detection in paddy leaves . A diverse dataset of paddy leaf images showing different types and severity levels of nutrient deficiencies is collected, and a Convolutional Neural Network (CNN) is used in order for image classification. The model is trained and tested on diverse dataset, demonstrating strong performance in accurately detecting nutrient deficiencies in paddy leaves.

Why it matches plant phenotyping methodsイネ葉画像から栄養欠乏という植物状態をCNNで分類する手法とデータセットが研究の中心であり、植物フェノタイピング手法に該当する。

abstractThe following study presents a novel approach to identifying nutrient deficiencies using neural networks and also provides a solution for their early detection in paddy leaves .

In-Field Nondestructive Detection of Nitrogen Status on 'Yotsuboshi' Strawberry Using Deep Learning Algorithm →

Nitrogen (N) management is critical for optimizing growth and fruit quality in open-field strawberry cultivation, demanding advanced technological solutions for reliable nutrient assessment. However, visual symptom diagnosis, though widely utilized for nutrient monitoring, is inherently subjective and prone to observer bias, resulting in inconsistent and often unreliable assessments. While available accurate tissue analysis is destructive and costly. Nondestructive, in-field imaging techniques such as the normalized difference vegetation index (NDVI) exist but require expensive multispectral imaging systems. To address these limitations, this study developed a streamlined methodology for in-field N status detection using deep learning on standard RGB images. The experiment utilized 'Yotsuboshi' strawberries in a randomized complete block design with sufficient nitrogen (T1) and deficient nitrogen (T2) treatments. To mitigate ambient light variability, a key challenge in open-field phenotyping, a low-cost phenotyping cylinder was developed for standardized smartphone image acquisition. Rigorous four-stage annotation criteria were also introduced to classify the nitrogen status in strawberry leaves as NormalN, LowN, or AdvancedLowN, ensuring a high-quality novel dataset. A YOLO11 model trained on this dataset achieved precision, recall, and mAP50 values exceeding 99%. Subsequent testing using the phenotyping cylinder yielded a mAP50 of 87%. In-field validation without a phenotyping cylinder also demonstrated robust performance under diffuse cloudy conditions (82.7% mAP50), outperforming direct sunlight (79% mAP50). Moreover, the model's classifications of 'NormalN' and 'LowN' statuses strongly corresponded with NDVI measurements, validating the accuracy of the RGB-based approach. This research demonstrates the significant potential of combining deep learning and phenotyping cylinder to create a rapid, low-cost, nondestructive and reliable tool for in-field nitrogen detection, with possible application across different crops and environmental conditions.

Why it matches plant phenotyping methodsイチゴ葉の窒素状態という植物状態を、RGB画像・深層学習・低コスト撮影筒で非破壊推定する手法を開発し、データセット作成と複数条件で検証しており、表現型取得・抽出法が研究の中心である。

abstractthis study developed a streamlined methodology for in-field N status detection using deep learning on standard RGB images.

Deep learning-based high-throughput phenotyping for tiller quantification in interspecific bentgrass hybrids using YOLOv8. →

Introduction: Tiller production is a critical determinant of turfgrass canopy density and plant performance, yet manual tiller counting is too labor-intensive for large breeding programs. Methods: To address this limitation, we evaluated 770 plants from an interspecific bentgrass hybrid population and developed three automated approaches for tiller quantification: a classical edge-based segmentation pipeline and two deep-learning models, Faster R-CNN and YOLOv8. Using a large annotated image dataset, we assessed each method's accuracy, robustness under occlusion, and computational efficiency. Results: Although two-stage detectors are often expected to provide superior precision for complex plant structures, the one-stage YOLOv8 model achieved the highest accuracy (R² = 0.97) and processed images substantially faster than Faster R-CNN, while both the edge-based method and Faster R-CNN showed reduced performance in dense canopies. Discussion: These findings demonstrate that recall-oriented one-stage detection can outperform more complex two-stage models for phenotyping tasks involving fine, highly occluded structures. The resulting workflow provides a reliable, high-throughput solution for generating biologically meaningful tiller counts and offers a transferable framework for integrating image-derived phenotypes into genetic analyses and breeding pipelines across grass species.

Why it matches plant phenotyping methodsイネ科植物の分げつ数を画像から自動抽出する複数手法を開発・比較検証しており、植物表現型取得法が研究の中心である。

abstractdeveloped three automated approaches for tiller quantification: a classical edge-based segmentation pipeline and two deep-learning models, Faster R-CNN and YOLOv8.

Tracking Recovery: Temporally-Matched 3D Gaussian Splatting of Ecosystems after Prescribed Burns →

Land management and stewardship teams continue to lack the tools to capture 3D spatiotemporal insights of the ecosystems they oversee. For wildfire management at the wildland-urban interface, teams face challenges in capturing vegetation growth over time after a fuel reduction program and connecting seasonal changes to the vegetation distribution across the treated area. Current approaches rely on triangle meshes or point clouds generated from photogrammetry or LiDAR surveys on drones or hiked traverses. However, the difficulties in optimizing these meshes lead to large triangles that inadequately approximate the bulk vegetation shape, and the point cloud data is often too sparse for local plant-scale understanding. To address this gap, we extend recent machine learning-based computer graphics techniques in 3D Gaussian Splatting (3DGS) to reconstruct scenes at the wildland-urban interface from handheld imagery with sufficient detail to identify species, capture individual leaves, recover plant stature, and disambiguate overhanging plant individuals. We develop a new method to match the 3DGS reconstruction of these scenes across months, associating plant growth across seasons interactively in 3D. To achieve the centimeter-level matching, we adapt the Umeyama algorithm and the iterative closest point algorithm from point cloud maps to the 3DGS scene, leveraging the probabilistic interpretation of the 3D Gaussian data structure and robustly handling visual and geometric changes associated with vegetation phenology over time. We have applied our method to recent pile burns at Stanford’s Jasper Ridge ’Ootchamin ’Ooyakma Biological Preserve at monthly intervals. We demonstrate differences in ecological response where some piles featured the unexpected return of a rare and threatened bushmallow, and others remained more barren. This pile burn microcosm implicates the need for plant-level 3D spatiotemporal models to understand ecosystem recovery to fire mitigation practices. Please visit the project page for the spatiotemporal alignment video and more information: https://danineamati.github.io/burn-ecorecovery.github.io/ Please visit the Stanford Data Repository for the field data: https://doi.org/10.25740/ws901xs0162

Why it matches plant phenotyping methods植物の3D画像再構成と時系列位置合わせ手法を開発し、植物の個体・葉・草丈・成長を抽出することが中心であるため、植物フェノタイピング手法研究に該当します。

abstractwe extend recent machine learning-based computer graphics techniques in 3D Gaussian Splatting (3DGS) to reconstruct scenes at the wildland-urban interface from handheld imagery with sufficient detail to identify species, capture individual leaves, recover plant stature, and disambiguate overhanging plant individuals.

An attention-based deep learning model for early detection of polyphagous shot hole borer infestations in plants. →

The Polyphagous Shot Hole Borer (PSHB) is a highly invasive beetle that has been spreading like an epidemic across agricultural and forestry landscapes in recent years. Its rapid and destructive spread has turned it into a major global threat, causing widespread damage that continues to grow with time. Countries like South Africa, the United States, and Australia have implemented extensive measures to control the spread of PSHB, including the establishment of specialized agricultural support centers for early detection. However, there is still a strong need to make PSHB detection more accessible, allowing even non-experts to easily identify infections at an early stage. Artificial Intelligence (AI) has shown great promise in plant disease detection, but a major challenge in the case of PSHB was the lack of a suitable dataset for training AI models. In the proposed work, we first created a dedicated dataset by collecting images of trees infected with PSHB. We applied a range of preprocessing techniques to refine the dataset and prepare it for AI applications. Building on this, we developed a novel AI-based method, where we trained a deep learning model using a multi-convolutional layer network combined with a Fourier transformation layer. Additionally, an attention mechanism and advanced feature extraction techniques were incorporated to further boost model performance. As a result, the proposed approach achieved an impressive top accuracy of 92.3% in detecting PSHB infections, showing the potential of AI to offer a simple, efficient, and highly accurate solution for early disease detection.

Why it matches plant phenotyping methods植物画像から感染状態を推定するデータセットと深層学習手法を開発し、性能評価まで行っており、病害フェノタイピング手法が中心である。

abstractwe first created a dedicated dataset by collecting images of trees infected with PSHB.

FCDNet: An Efficient and Cost-Effective Strawberry Disease Detection Model for Smart Farming Management. →

With the rapid development of precision agriculture and smart farming management, accurate crop disease detection has become a critical tool for optimizing agricultural resource allocation, controlling operational costs, and supporting scientific plant protection strategies. However, real-world field environments are often characterized by strong background interference, multiple concurrent diseases, and fine-grained lesion differences, posing significant challenges to existing detection methods in practical agricultural Internet of Things (IoT) applications. In this paper, we propose Freq-spatial Context Dynamic Network(FCDNet), an efficient and cost-effective detection model tailored for multi-category strawberry disease recognition in complex field management scenarios. The proposed model integrates a Freq-Spatial Feature Module (FSFM), a Context Guide Fusion Module (CGFM), and a Task Align Dynamic Detection Head (TADDH), enabling enhanced expression of high-frequency micro-lesions, adaptive filtering of field background noise, and spatial alignment of classification and regression tasks, while maintaining a lightweight architecture suitable for low-cost agricultural edge devices. Extensive experiments conducted on the newly constructed Strawberry Disease Dataset-7(S7DD) demonstrate that FCDNet consistently outperforms existing mainstream methods, achieving an F1-score of 91.0% and an mAP@0.5 of 94.6%. The model's architectural robustness and capacity for generalization are further substantiated by evaluations across diverse agricultural datasets using PlantDoc and ALDOD. Ultimately, FCDNet became a practical and cost-effective tool for real-time detection of strawberry diseases, directly supporting more accurate yield forecasting and risk management in smart agriculture systems.

Why it matches plant phenotyping methodsイチゴ葉の病斑・病害状態を画像から認識するモデルを開発し、複数データセットで性能評価しているため、植物病害フェノタイピング手法が中心である。

abstractwe propose Freq-spatial Context Dynamic Network(FCDNet), an efficient and cost-effective detection model tailored for multi-category strawberry disease recognition

Multimodal AI for plant stress classification across multiple plant species →

Plant stress monitoring is invaluable in realizing sustainable agriculture because it enables the people practicing it to take early measures to counteract losses in yield caused by environmental stressors like drought and nutrient deficiencies, as well as caused by pathogen infections. The proposed study presents a new Multi-Modal Vision Transformer (MMViT) architecture that is designed to combine both thermal and RGB imagery to take detection to the next level. In order to support this methodology and further studies, we are now publicly releasing a new collection of synchronized thermo-RGB image pairs of stressed and healthy plants, collected both in controlled settings and in the field. The data is labeled to differentiate various stress phenotype and contains over 4286 of images, and hence forms a substantial platform to evaluate multimodal plant phenotyping methods. Empirical evaluations indicate that MMViT model achieves a general classification of 94.3% when using the two modalities, which is better than the single-modality ViT used on the thermal images (85.5%) and the RGB images (93.3%). These experimental results emphasize the performance of multimodal fusion whereby the other spectral cues are used to complement a stress classification. The described framework, together with the useful dataset, will contribute to the advancement of precision agriculture as it is an open and data-driven instrument to monitor plant health automatically.

Why it matches plant phenotyping methods熱画像とRGB画像を統合して植物ストレス表現型を分類するモデルを開発し、公開データセットと性能評価も提示しており、植物フェノタイピング手法が中心である。

abstractThe proposed study presents a new Multi-Modal Vision Transformer (MMViT) architecture that is designed to combine both thermal and RGB imagery to take detection to the next level.

ICAG ‐Net: An Interactive CNN –Transformer Architecture With Attention‐Guided Gated Fusion for Crop Disease Detection →

ABSTRACT Smart agriculture based on the use of Artificial Intelligence for crop disease detection to ensure food security. Fungal disease is one of the major causes that affects the quality of the vegetables. Convolutional neural networks (CNNs) and vision transformers (ViTs) enable the detection of crop diseases at an early stage, allowing farmers to take preventive measures and minimize further losses. The proposed method introduces an ensemble of Custom CNN to understand multilevel local features and Pretrained ViT to capture global dependencies as well as contextual information from the dataset. An interactive cross attention module (ICAM) facilitates bidirectional information exchange between CNN and transformer token representations, while an attention guided gated fusion (AGGF) mechanism adaptively combines complementary features. The Indian Crop Visual Disease Dataset (ICVDD‐5) has been developed in a real field for the proposed work with the help of domain experts. The dataset contains 880 diverse images depicting both healthy and diseased specimens of five vegetable crops. The crops selected for this research initiative include Brinjal, Cabbage, Chili, Okra, and Tomato. These five crops are examined for about 21 distinct disease classes. Comprehensive ablation studies are conducted to prove the contributions of each architectural component, including CNN‐only, ICAM‐disabled, and AGGF‐disabled configurations. Experimental results demonstrate that the proposed ICAG‐Net achieves a test accuracy of approximately 70%–73% with improved macro‐F1 score compared to baseline CNN models under identical training settings. The novelty of this work lies in an extensible solution for real world crop disease diagnosis systems and offers insights into hybrid CNN–transformer architectures for small scale agricultural datasets.

Why it matches plant phenotyping methodsCNN・Transformerによる植物病害症状の画像検出手法を開発し、実圃場画像データセットの構築、アブレーション、ベースライン比較で性能検証しているため、植物フェノタイピング手法が中心である。

abstractThe proposed method introduces an ensemble of Custom CNN to understand multilevel local features and Pretrained ViT to capture global dependencies as well as contextual information from the dataset.

Mobile-assisted deep learning framework for identification of insect pests and diseases of maize from field images. →

Maize ( Zea mays L. ) production is severely affected by diseases and insect pests, leading to significant yield losses when timely diagnosis and management interventions are not implemented. Although automated image-based diagnostic systems have shown promising results, most existing studies address diseases or pests independently, rely on controlled datasets, and offer limited robustness under real field conditions. To address these limitations, this study proposes a unified deep learning-based framework for integrated identification of maize diseases and insect pests under natural field environments by combining object detection and image classification within a mobile-assisted diagnostic system. Four economically important diseases and insect pests were investigated: Maydis Leaf Blight (MLB), Turcicum Leaf Blight (TLB), Common Rust, and Fall Armyworm (FAW). MLB and TLB were addressed using YOLO-based object detection architectures, while Common Rust and FAW were treated as image-level classification tasks using lightweight deep learning models optimised for mobile inference. A self-collected dataset comprising 10,343 images across four classes was acquired under real field conditions to capture variability in background complexity, illumination, phenological stages, and symptom expression. Experimental results on an independent test set comprising original images demonstrate that MobileViT achieved the highest classification accuracy (99%) for image-level disease and pest recognition, whereas YOLOv11n outperformed other detection models, achieving the best performance for MLB and TLB lesion detection with mAP@0.5 of 0.875. Grad-CAM-based visual explanation analysis confirmed that the classification models focused on disease lesions and pest-infested regions, supporting interpretability. The framework was successfully deployed via a mobile application, enabling image acquisition, automated validation, diagnosis, and the generation of management recommendations. The results highlight the accuracy, robustness, and operational feasibility of the proposed system for in-field diagnosis of maize diseases and insect pests, supporting early detection and sustainable crop protection.

Why it matches plant phenotyping methodsトウモロコシ葉の病斑と害虫被害領域を圃場画像から検出・分類する画像解析手法と、モバイル診断プラットフォームを開発・評価しており、植物の病害状態の取得が中心的です。

abstractthis study proposes a unified deep learning-based framework for integrated identification of maize diseases and insect pests under natural field environments by combining object detection and image classification within a mobile-assisted diagnostic system.

Multi‐Spectral Gaussian Splatting with Neural Color Representation →

Abstract 3D Gaussian Splatting (3DGS) [KKLD23] has transformed novel‐view synthesis from RGB images, yet remains restricted to the visible spectrum. Many applications, including agricultural monitoring, rely on multi‐spectral imaging, where spectral camera alignment and scalability pose major challenges. We present MS‐Splatting—a multi‐spectral 3DGS framework enabling unified multi‐view consistent reconstruction and rendering across both visible and invisible spectra. Our key component is a neural color representation that encodes per‐primitive features shared across spectral bands, decoded through a shallow multi‐layer perceptron into spectrum‐specific radiance. By leveraging inter‐band correlations, this formulation enhances detail while reducing memory consumption compared to independent band modeling via per‐channel modeling with spherical harmonics. Our method enables accurate parallax‐free novel‐view vegetation index rendering for plant monitoring and enhances RGB novel view synthesis quality by exploiting details revealed through multi‐spectral bands. Our evaluation demonstrates that MS‐Splatting exceeds the current leading methods in both categories. In addition, we introduce a multi‐spectral dataset from aerial captures covering outdoor environments, specifically designed for evaluating these applications. We will release our code and dataset to facilitate further research. The project page is located at: https://meyerls.github.io/ms_splatting

Why it matches plant phenotyping methodsマルチスペクトル3D再構成法を開発し、植物モニタリング用の植生指数レンダリングを実現することが中心で、評価用データセットも提供している。

abstractWe present MS‐Splatting—a multi‐spectral 3DGS framework enabling unified multi‐view consistent reconstruction and rendering across both visible and invisible spectra.

SAB-DeepLabV3+: A Semantic Segmentation Framework for Mapping Maize Waterlogging from Single-Date Multispectral Imagery →

Rapid identification of maize waterlogging is essential for post-disaster agricultural assessment, but most existing methods rely on multi-temporal imagery that is often unavailable immediately after extreme rainfall events. This study proposes SAB-DeepLabV3+, a semantic segmentation model for mapping waterlogged maize from single-date multispectral imagery within pre-extracted maize planting areas. Built on DeepLabV3+, the model integrates three task-specific modules: a Spectral-Spatial Information Enhancement Module to improve feature discrimination under spectral mixing, an Adaptive Multi-Scale Pooling Module to capture heterogeneous patch sizes, and a Boundary Enhancement Module to refine transition zones. A pixel-level dataset containing 12,198 image patches was constructed from 62 multispectral scenes collected across five major maize-producing cities in Heilongjiang Province, China, during 2022–2024. On the test set, SAB-DeepLabV3+ achieved a waterlogged-class IoU of 68.30%, mIoU of 80.37%, mF1 of 88.62%, and OA of 93.49%, outperforming DeepLabV3+. Leave-one-city-out evaluation further produced an average mIoU of 76.56% and a waterlogged-class IoU of 63.45%. These results indicate that single-date high-resolution multispectral imagery can support rapid and reliable maize waterlogging mapping.

Why it matches plant phenotyping methodsマルチスペクトル画像からトウモロコシの水害状態を抽出するセマンティックセグメンテーション手法を開発し、データセットと都市間評価で検証しているため、植物フェノタイピング手法が中心である。

abstractThis study proposes SAB-DeepLabV3+, a semantic segmentation model for mapping waterlogged maize from single-date multispectral imagery within pre-extracted maize planting areas.

Hybrid CNN-transformer architecture for multi-class crop disease detection and severity assessment: CropHybrid-Net with benchmark evaluation on CropDisease-12 →

Grain diseases lead to losses of 20-40% of the harvests each year, representing a threat to the food security of the world. Accurate and automated diagnosis of disease from remote picture taking would be key to prompt and directed interventions. Most current deep-learning approaches, however, are based on controlled lab images, on a single crop and ignore the assessment of disease severity. CropHybrid-Net is a three-branch ensemble architecture with ResNet-50, EfficientNet-B4 and Swin Transformer (Swin-T) that combines them using Efficient Channel Attention (ECA) fusion layer for simultaneous disease detection and severity estimation. CropDisease-12 is a benchmark dataset of 43200 images belonging to 12 classes representing four major crops (tomato, wheat, rice, and cotton) from PlantVillage and its own disease dataset collected in Yavatmal, Maharashtra, India. When evaluated on the CropDisease-12 test split, CropHybrid-Net outperforms all baselines tested such as standalone Swin-T (94.5%), ViT-B/16 (93.9%) and EfficientNet-B4 (93.7%), with the highest accuracy of 97.8%, and macro f1 score of 97.3%. The average value of AUC for the 12 classes is 0.995. In addition, a comprehensive literature review has been conducted, comprising of 62 papers (2015-2024), and grouped into five research streams: conventional machine learning, CNN-based methods, transfer learning, transformer-based methods, and multi-task severity approaches. The Grad-CAM visualizations are in line with the locations of biologically meaningful lesions. The framework proposed is deployed on common precision agriculture-edge of-use devices and achieves the goal of 39.3 ms per image, being relevant to smart precision agriculture applications.

Why it matches plant phenotyping methods植物病害の画像から病害状態と重症度を推定するCNN・Transformer手法を開発し、ベンチマークデータセットで評価しているため、植物フェノタイピング手法が中心である。

abstractCropHybrid-Net is a three-branch ensemble architecture with ResNet-50, EfficientNet-B4 and Swin Transformer (Swin-T) that combines them using Efficient Channel Attention (ECA) fusion layer for simultaneous disease detection and severity estimation.

Data-driven estimation of lettuce biophysical traits using multidimensional sensing →

• Multisensor platform integrates RGB, depth, IR, and RTK-GPS data streams • Automated plant segmentation and 3D reconstruction extract plant traits in field conditions • System validation shows high correlation with manual and lab measurements • Public RGB-D lettuce dataset released to support reproducible AI research Accurate monitoring of leafy vegetable crops is essential to evaluate plant health, growth, yield, and quality, yet conventional methods based on manual measurements are labor-intensive and error-prone. This study proposes a data-driven framework for automated in-field monitoring of a lettuce crop based on multidimensional data acquired by a ground platform under various field conditions. Specifically, an advanced perception system is developed, including imaging and localization sensors to capture high-resolution visual, structural, and georeferenced information on the crop. An image processing pipeline is then proposed using zero-shot learning for plant segmentation, followed by 3D phenotyping techniques based upon computational geometry to automatically estimate plant biophysical traits, thus minimizing human input. An experimental trial conducted in a test field in Bari, Italy, between April and May 2025 validated the approach against manual and laboratory estimations. The results demonstrate strong correspondence between automated and reference measurements with a Pearson correlation coefficient r > 0.9 for key traits, confirming the potential of the framework. The influence of different nitrogen levels on the growing cycle is also evaluated, showing that the proposed system may provide a useful tool for decision support in lettuce crop monitoring and management.

Why it matches plant phenotyping methodsRGB・深度・IR等を統合したセンシング、植物セグメンテーション、3D形状解析による形質推定を開発し、手測定・実験室測定で検証しているため、フェノタイピング手法が中心である。

abstractAn image processing pipeline is then proposed using zero-shot learning for plant segmentation, followed by 3D phenotyping techniques based upon computational geometry to automatically estimate plant biophysical traits

A paradigm-level evaluation of deep learning architectures for plant pest and disease recognition under data-scarce field conditions →

Introduction Accurate recognition of plant pests and diseases under field conditions remains challenging due to complex symptom morphology, environmental variability, and limited annotated data. While deep learning has been widely adopted for image-based diagnosis, existing studies are often model-centric and evaluated under heterogeneous experimental settings, making it difficult to derive paradigm-level insights into data efficiency, robustness, and practical deployment suitability. Methods In this research, we systematically compare three main types of vision models: Convolutional Neural Networks (CNNs), Vision Transformers, and mixed State Space Model-based structures (MambaVision). These models are applied to classify images of pests and diseases across multiple crops. Using a unified and reproducible experimental framework, we benchmark representative models across multiple training regimes, diverse crop species, and symptom types reflecting realistic agricultural scenarios. Results Results reveal clear paradigm-level differences. CNN-based models perform competitively on diseases dominated by localized lesion textures but show limited robustness for symptoms requiring global spatial interpretation. Transformer-based models benefit from global dependency modeling yet exhibit increased instability under small-sample conditions. In contrast, hybrid MambaVision-based models consistently demonstrate superior data efficiency and robustness, retaining approximately 60–80% accuracy under extreme data scarcity (1% training samples) and achieving stable, high F1-scores across symptom types that require joint modeling of fine-grained textures and long-range spatial distribution. Furthermore, performance–efficiency analysis shows that hybrid MambaVision-based models achieve a more favorable accuracy–computational cost trade-off than CNN-based and Transformer-based models, supporting deployment under practical resource constraints. Discussion Overall, this study provides pathology-oriented and deployment-aware insights into how architectural inductive bias interacts with symptom morphology and data availability, highlighting hybrid MambaVision-based models as a robust and effective solution for real-world plant pest and disease recognition.

Why it matches plant phenotyping methods植物病害の症状画像を対象に、複数の深層学習アーキテクチャを統一条件で比較・ベンチマークしており、植物の病害状態を推定する方法の評価が研究の中心である。

abstractUsing a unified and reproducible experimental framework, we benchmark representative models across multiple training regimes, diverse crop species, and symptom types reflecting realistic agricultural scenarios.

PLDC-Net: A Domain-Specific Base Model for Plant Leaf Disease Classification Domain Adaptation Tasks. →

Plant diseases are the cause of heavy losses of crop production and, therefore, a big contributor to food shortages. Identifying these diseases as early as possible is important to limit the negative effects that these diseases have on the yields, as slow response time will lead to the spread of diseases and further loss. Traditionally, trained staff will go into the fields, multiple times during the growth period, and inspect the plants in samples through field disease monitoring. These traditional processes are time-consuming and costly, and can be error-prone, if the staff is not properly educated or if the staff simply makes mistakes due to oversight, for example. To aid farmers with the process of correctly identifying diseases, artificial intelligence deep learning methods have been employed in recent years. However, to train such deep learning models, one needs to obtain sufficiently large and high-quality datasets and a model architecture that is capable of extracting relevant features to accurately classify the plant leaves. Datasets are still a limitation in the field of plant leaf disease classification. As such, domain adaptation methods such as transfer learning are often employed to overcome this data shortage. However, in current research, these domain adaptation methods almost exclusively rely on ImageNet as the pretraining dataset, a dataset that is domain unrelated to plant leaf disease detection, and models are often left unmodified and un-optimized as a result. In this work, we propose the pretraining of an improved attention-based and SiLU-activated DenseNet201 architecture called PLDC-Net that is pretrained on a large-scale plant leaf disease dataset constructed by the authors to create a domain-specific base model for better domain adaptation to new plants and diseases, validating the improved results through transfer learning, fine-tuning, one-shot learning, and few-shot learning. PLDC-Net has managed up to just over 24% improvements in F1-Score over the baseline in domain adaptation results.

Why it matches plant phenotyping methods植物葉の病害状態を画像分類するモデルと専用データセットを開発し、転移学習等で性能検証しており、表現型取得・推定手法が研究の中心である。

abstractwe propose the pretraining of an improved attention-based and SiLU-activated DenseNet201 architecture called PLDC-Net that is pretrained on a large-scale plant leaf disease dataset constructed by the authors to create a domain-specific base model

CocoaDetectDB: A TinyML-Oriented Image Dataset for Cocoa Plant Disease Detection →

The application of computer vision in precision agriculture has demonstrated considerable promise in automated plant disease detection. However, the effectiveness of such approaches is strongly dependent on the availability of high-quality, domain-specific datasets, particularly for deployment on resource-constrained edge devices. This paper introduces CocoaDetectDB, a publicly available image dataset developed for the detection of cocoa plant diseases under Tiny Machine Learning (TinyML) constraints. The dataset comprises images of healthy cocoa pods and three major cocoa diseases—Cocoa Black Pod Disease (CBD), Cocoa Swollen Shoot Virus Disease (CSSVD), and Frosty Pod Rot (FPR)—captured under real-world field conditions and supplemented with openly accessible public data. Images were curated, cleaned, and resized to a uniform resolution of 112 × 112 pixels to support low-memory and low-power inference. To validate the suitability of the dataset for automated disease classification, baseline experiments were conducted using MobileNetV2 and a lightweight quantized TensorFlow Lite model. Experimental results demonstrate classification accuracies of 99.13% and 93.75%, respectively, indicating that CocoaDetectDB contains sufficiently discriminative features for both conventional lightweight models and TinyML deployment. The dataset is intended to support future research in cocoa disease detection, edge AI, and resource-efficient agricultural monitoring systems.

Why it matches plant phenotyping methodsココア植物の病害状態を画像で判定する公開データセットを構築し、軽量モデルで適合性を検証しており、画像ベースの植物表現型取得・分類が中心である。

abstractThis paper introduces CocoaDetectDB, a publicly available image dataset developed for the detection of cocoa plant diseases under Tiny Machine Learning (TinyML) constraints.

An End-to-End Precision Phenotyping Framework: Rice Panicle Detection and Counting in Complex Fields via Lightweight DETR →

Accurate, high-throughput quantification of rice panicles plays a vital role in advancing precision yield prediction. However, transitioning to real-time, edge-deployable unmanned aerial vehicle phenotyping is often impeded by extreme spatial scale variations from altitude fluctuations and complex unstructured backgrounds. To address this, we constructed a comprehensive composite dataset specifically capturing multi-altitude and varying illumination field conditions. We then propose Panicle-DETR, a highly optimized precision phenotyping framework incorporating a frequency-aware CSP backbone. By projecting visual perception into the frequency domain, the architecture inherently suppresses low-frequency environmental noise and minimizes computational redundancy. Furthermore, a Lossless Feature Encoder prevents the irreversible pixel decimation of micro-targets across varying operational altitudes, while a composite metric loss explicitly disentangles heavily adhered panicle clusters. Evaluated on our composite dataset, Panicle-DETR achieved an outstanding detection Precision of 90.97% alongside robust agronomic counting stability, demonstrated by a Mean Absolute Error of 4.28 and an \( R^2 \) of 0.957. With a compact footprint of only 13.78 M parameters, this framework fundamentally overcomes the computational and spatial limitations of traditional vision models, establishing a highly reliable paradigm for autonomous, onboard agricultural monitoring.

Why it matches plant phenotyping methods米の穂を画像から検出・計数する軽量なUAV表現型解析フレームワークを開発し、複合データセット上で性能評価しており、表現型取得・抽出手法が中心である。

abstractwe constructed a comprehensive composite dataset specifically capturing multi-altitude and varying illumination field conditions.

SynerFANet: a synergistic hybrid architecture for advanced plant leaf disease detection →

Accurate identification of plant leaf diseases is essential for modern agriculture. This paper presents SynerFANet, a hybrid deep learning framework designed to improve disease detection in sugarcane leaves. SynerFANet integrates two core modules: AdaptiveMBNet and WiseAttentionNet for comprehensive feature extraction and processing. AdaptiveMBNet combines MBConv layers with attention mechanisms to improve feature quality and reduce computation, enabling more accurate disease detection. WiseAttentionNet incorporates attention mechanisms into depthwise and expansion layers to enhance feature recalibration. The combination of the two cores inside SynerFANet can improve the representation capacity and robustness of the overall model. We evaluate the model using two datasets: a new proposed field-collected SugarLeaf-IDN dataset and the publicly available PlantVillage dataset. SynerFANet achieves superior accuracy with a moderate parameter size and GFLOPs, providing a balanced trade-off between predictive performance and computational cost, and exhibiting stable convergence during training. SynerFANet achieves 95.81% validation accuracy on our challenging real-world SugarLeaf-IDN dataset and 99.85% (SOTA) on the controlled PlantVillage benchmark.

Why it matches plant phenotyping methodsサトウキビ葉の病徴を画像から検出する深層学習モデルを開発し、実フィールドおよび公開データセットで性能評価しており、植物表現型取得・判定手法が中心である。

abstractThis paper presents SynerFANet, a hybrid deep learning framework designed to improve disease detection in sugarcane leaves.

Precise identification of tomato leaf diseases based on the real background of DPEN. →

As an important economic crop, tomato is vulnerable to various diseases, and these diseases often have high visual similarity, making identification difficult. A delay in diagnosis can have a significant effect on tomato yields. Traditional manual visual inspection methods have poor accuracy, while laboratory diagnostic methods are inefficient, making them unsuitable for large-scale agricultural scenarios. To address this challenge, this study involved the collection and construction of a tomato leaf disease dataset in a real planting environment, and AutoAugment was used to achieve sample diversity and balance the number of training samples in different categories. Additionally, a dual-path ensemble network (DPEN) was proposed, which combines the multiscale feature extraction advantages of GoogLeNet with the dense connection mechanism of DenseNet121. The experimental results show that, compared with the comparison models, the DPEN achieves an identification precision of 98.80% on the self-built dataset, which is an improvement of 2.33% to 9.24%, and a reduction in the number of parameters by 7.09 M compared with GoogLeNet and 2.07 M compared with DenseNet121. The experimental results on public datasets further demonstrated the accuracy of the proposed DPEN model in identifying tomato leaf diseases in complex backgrounds. These results prove that the DPEN model can achieve precise, rapid, and efficient identification of tomato leaf diseases in complex backgrounds, providing technical support for smart agriculture applications.

Why it matches plant phenotyping methodsトマト葉の病徴を画像から識別する深層学習手法を開発し、実環境データセットと公開データセットで精度検証しているため、植物病害状態のフェノタイピング手法が中心である。

abstracta dual-path ensemble network (DPEN) was proposed

From UAV Imagery to Agronomic Reasoning: A Multimodal LLM Benchmark for Plant Phenotyping →

To improve crop genetics, high-throughput, effective and comprehensive phenotyping is a critical prerequisite. While such tasks were traditionally performed manually, recent advances in multimodal foundation models, especially in vision-language models (VLMs), have enabled more automated and robust phenotypic analysis. However, plant science remains a particularly challenging domain for foundation models because it requires domain-specific knowledge, fine-grained visual interpretation, and complex biological and agronomic reasoning. To address this gap, we develop PlantXpert, an evidence-grounded multimodal reasoning benchmark for soybean and cotton phenotyping. Our benchmark provides a structured and reproducible framework for agronomic adaptation of VLMs, and enables controlled comparison between base models and their domain-adapted counterparts. We constructed a dataset comprising 385 digital images and more than 3,000 benchmark samples spanning key plant science domains including disease, pest control, weed management, and yield. The benchmark can assess diverse capabilities including visual expertise, quantitative reasoning, and multi-step agronomic reasoning. A total of 11 state-of-the-art VLMs were evaluated. The results indicate that task-specific fine-tuning leads to substantial improvement in accuracy, with models such as Qwen3-VL-4B and Qwen3-VL-30B achieving up to 78%. At the same time, gains from model scaling diminish beyond a certain capacity, generalization across soybean and cotton remains uneven, and quantitative as well as biologically grounded reasoning continue to pose substantial challenges. These findings suggest that PlantXpert can serve as a foundation for assessing evidence-grounded agronomic reasoning and for advancing multimodal model development in plant science.

Why it matches plant phenotyping methodsPlantXpertは作物フェノタイピング向けの画像ベンチマークとVLM評価基盤を構築しており、表現型解析手法・データセットの開発が研究の中心である。

abstractwe develop PlantXpert, an evidence-grounded multimodal reasoning benchmark for soybean and cotton phenotyping.

Deep learning-based estimation of plant functional traits from canopy spectral reflectance. →

Large-scale canopy-level plant trait quantification enhances crop yield and quality assessment, supports sustainable forestry economic development, and improves ecosystem monitoring globally. However, traditional methods relying on vegetation spectral libraries and machine learning models often face challenges in capturing the nonlinear and multivariate characteristics of canopy spectral responses. To overcome these challenges, we propose a novel deep learning framework, the Canopy-level Plant Functional Trait Retrieval Network (CTRN), which integrates Kolmogorov-Arnold Networks (KAN), Transformer, and Convolutional Neural Networks (CNN) to effectively extract informative representations from high-dimensional hyperspectral reflectance. The model is trained and evaluated using a comprehensive spectral-trait dataset, covering various plant species, different sensors, and multiple continents, and focuses on ten key canopy functional traits. Experimental results show that CTRN consistently outperforms other models, achieving R 2 values greater than 0.82 across all traits. Furthermore, even with just 44 spectral bands at a 40 nm resolution, CTRN demonstrates commendable accuracy in estimating LMA and C, with R 2 values approaching 0.80. These findings highlight the robust ability of the model to characterize complex associations between canopy spectra and plant functional traits, supporting accurate parameter retrieval in ecological and agricultural applications.

Why it matches plant phenotyping methodsキャノピーのハイパースペクトル反射から植物機能形質を推定する深層学習手法を開発・評価しており、形質取得法が研究の中心である。

abstractwe propose a novel deep learning framework, the Canopy-level Plant Functional Trait Retrieval Network (CTRN)

Scalable AI-Driven Maize Plant Disease Detection using Convolutional Neural Network →

Plant diseases have a significant impact on global food security, especially in staple crops like maize (Zea mays). Traditional disease detection systems depend on professional visual inspection, which is labor-intensive, time-consuming, and not scalable for large agricultural areas. Convolutional Neural Networks (CNNs) are used in this study's deep learning (DL) architecture to detect maize leaf diseases accurately and automatically. A curated dataset of approximately 7,000 high-resolution maize leaf photos was created, representing four classes: healthy, Common Rust (Puccinia sorghi), Northern Leaf Blight (Exserohilum turcicum), and Gray Leaf Spot (Cercospora zeae-maydis). Data were sourced from the Plant Village dataset, real-world field collections from Indian farms, and supplemented synthetically to simulate varied climatic circumstances. Advanced methods including as adaptive learning rate scheduling, gradient clipping, and significant data augmentation were used to train a bespoke CNN model that was improved by transfer learning with ResNet50 and VGG16 backbones. The model attained a test accuracy of 98.2%, beating classic machine learning algorithms like SVM (88.5%) and Random Forest (84.3%). Visualization approaches such as feature maps, Grad-CAM, and LIME improved interpretability and showed the model's capacity to locate disease-relevant features. Web-based user engagement is made possible by deployment-ready implementation, which enables farmers to upload leaf photos for immediate diagnosis. With the potential to cut maize crop losses by 20–30%, this research offers a scalable and affordable alternative to early disease detection in precision agriculture. Future research will investigate autonomous farm management with drone-based real-time surveillance and IoT system integration.

Why it matches plant phenotyping methodsトウモロコシ葉の病害状態を画像から推定するCNN手法を開発し、データセット、比較評価、精度検証、解釈性分析まで行っており、植物表現型取得が中心である。

abstractConvolutional Neural Networks (CNNs) are used in this study's deep learning (DL) architecture to detect maize leaf diseases accurately and automatically.

Description of morphological characteristics of wheat spike as a digital certificate in the SpikeDroidDB database. →

It has been repeatedly shown that spike productivity is the main component of wheat yield. The main spike parameters related to productivity are size, the number of grains and spikelets per spike, and the presence or absence of awns. In modern genetic research, morphometric analysis of hundreds and thousands of spikes is required to determine the loci that control spike productivity traits. On the other hand, thousands of accessions in modern collections of wheat genetic resources need detailed description. These considerations motivate the development of digital technologies for describing spike traits in wheat, which can be achieved through image analysis methods. These methods allow for automated acquisition of trait values that can serve as the basis for digital plant collections. Here we propose an extended set of spike characteristics obtained both manually and through digital image analysis and present plant characterization. These data form the basis of the updated version of the SpikeDroidDB database (http://spikedroid.biores.cytogen.ru/). The digital description of the spike consists of two blocks. The block of uploaded data includes a description of the plant and contains five tables: collection; variety sample (year of cultivation (vegetation), sowing identifier, taxonomic information, etc.), planting site, and characteristics of the spike determined manually (length, width of frontal and lateral views, type and color of the spike, etc.) The block of extracted features includes spike characteristics obtained by digital phenotyping and contains six tables: characteristics of the spike outline in the image; characteristics of the quadrangle model, values of the color components of the spike, dominant colors of the spike, and texture characteristics of the spike in the image. The most illustrative and significant features of the spike have been identified, allowing for the formation of the spike digital certificate, which includes size, shape, and color features derived from the digital images. The features forming the digital certificate have been compared between two wheat species, T. aethiopicum and T. carthlicum. It is shown that the features of the digital certificate allow for a clear representation of the spike model and the identification of distinct parameters: colors of the spike and awns and roundness of the frontal view of the spike. The database interface has been supplemented with the ability to upload data on plant and spike characteristics, as well as their images, in the batch mode.

Why it matches plant phenotyping methods小麦穂の画像解析による形態形質の自動取得、デジタル表現、データベース基盤を中心に開発・提示しており、植物フェノタイピング手法が研究の中核である。

abstractThese considerations motivate the development of digital technologies for describing spike traits in wheat, which can be achieved through image analysis methods.

Deep Learning-Based Plant Disease Prediction Using Real-Field Image Data →

Abstract— Plant diseases remain a persistent threat to global agricultural output, annually destroying an estimated 20–40% of total crop yields worldwide. Prompt and reliable identification of infection at early stages is indispensable for guiding timely intervention and protecting food security. Conventional diagnostic workflows, which depend on field visits by trained agronomists, are inherently subjective, time-consuming, and impractical at the scale of modern large-area farming. To address these operational gaps, this study proposes a fully automated deep learning pipeline tailored for classifying plant diseases from photographs collected under uncontrolled, realistic field conditions. The core of the architecture is a hybrid model that couples EfficientNet-B4 [3] with a Convolutional Block Attention Module (CBAM) [4], equipping the network with the capacity to localize and emphasize abnormal leaf tissue while filtering out irrelevant scene elements. The system is developed and benchmarked on a purpose-built dataset of 54,306 labeled images representing 26 disease classes across 14 crop species. A structured preprocessing workflow—encompassing Contrast Limited Adaptive Histogram Equalization (CLAHE), mosaic-based class balancing, and mixup regularization [10]—is incorporated to enhance the model's tolerance to lighting inconsistencies and skewed class distributions. On the held-out test partition, the proposed model attains a top-1 accuracy of 96.7%, outperforming six well-established CNN baselines. Gradient-weighted Class Activation Mapping (Grad-CAM) [11] visualizations further confirm that the model directs its attention toward pathologically relevant regions, lending credibility to its predictions. These properties collectively position the framework as a strong foundation for lightweight, smartphone-deployable disease advisory tools for smallholder farmers. Keywords—Plant Disease Detection, Deep Learning, EfficientNet, CBAM Attention, Grad-CAM, Real-Field Image Dataset, Convolutional Neural Networks, Precision Agriculture

Why it matches plant phenotyping methods植物画像から病害状態を推定する深層学習パイプラインを開発し、専用データセットとベースライン比較で検証しており、病害表現型の取得・推定が中心である。

abstractthis study proposes a fully automated deep learning pipeline tailored for classifying plant diseases from photographs collected under uncontrolled, realistic field conditions.

A wheat seedling detection model based on efficient feature extraction and coordinate attention mechanism →

Accurate detection of wheat seedlings is crucial for monitoring early population establishment and evaluating sowing quality. However, detection in real field environments remains challenging due to diverse seedling morphology, varying planting densities, occlusion, and complex background interference. Although deep learning has promoted the development of agricultural vision systems, existing wheat seedling detection methods still suffer from two key limitations: (1) insufficient modeling of spatial contextual relationships, leading to degraded accuracy under dense planting and complex field conditions; and (2) difficulty in balancing detection performance and computational efficiency, restricting real-time deployment on resource-limited agricultural devices. To address these issues, this study proposes Transformer-Coordinate Attention-Efficient YOLO (TCE-YOLO), a detection framework designed with three key modules: (1) the Depthwise-Transformer-Vision (DTV) module integrates Depthwise Separable Convolutions (DSC), Vision Transformer, and multi-scale spatial pooling to efficiently represent local structures, spatial context, and global patterns of wheat seedlings; (2) the Feature Enhancement Module(FEM) incorporates coordinate attention to enhance seedling-related features while suppressing background interference; and (3) the Feature Coordination Module (FCM) performs multi-scale feature interaction with reduced computational cost. These components jointly improve robustness under dense planting and complex field conditions while maintaining lightweight deployment characteristics. Furthermore, we construct the Wheat Seedling Dataset (WSD), covering multiple planting densities, varieties, and field environments across two growing seasons. Experimental results show that TCE-YOLO outperforms mainstream detectors while maintaining high efficiency, providing a deployable solution for wheat seedling detection under real field conditions.

Why it matches plant phenotyping methods小麦苗の画像検出手法を開発し、複数条件・作期を含むデータセットを構築して性能比較しているため、植物フェノタイピング手法が研究の中心である。

abstractthis study proposes Transformer-Coordinate Attention-Efficient YOLO (TCE-YOLO), a detection framework designed with three key modules

SorghumBase: a knowledgebase for sorghum genomics, phenomics, and stakeholder engagement. →

Centralizing valuable community data and resources into a user-friendly interface and accessible repository has become essential for agricultural science; embracing Findable Accessible, Interoperable, and Reusable (FAIR) principles is now standard for effective databases. SorghumBase (https://www.sorghumbase.org) is a knowledgebase designed for the sorghum research community. The SorghumBase team curates genomic, transcriptomic, variation, and phenotypic information and aggregates community events, providing rich visualizations and bulk data access. The modular framework of the database is built with open-access software to yield a robust, modifiable, and sustainable data infrastructure. Release 9 of SorghumBase includes: (i) 88 sorghum reference genomes and an updated pan-gene index, (ii) over 100 million variants have been mapped onto the 2 genomes, BTx623 and Tx2783, (iii) assignment of 41 million Reference Cluster SNP identifiers (rsIDs) from BTx623 across the pan-genome, (iv) updated gene search homology, gene expression, and germplasm visualizations and features, (v) added and standardized 234 phenotypic data from 40 community-generated GWAS studies and 148 traits from the Sorghum QTL Atlas (Oz Sorghum), (vi) improved news, funding, and a research content management system for community access and interaction, (vii) outreach materials including training documents and videos, and (viii) community engagement initiatives through training and working groups. SorghumBase serves as a hub for sorghum data and stakeholder engagement while promoting community standards to drive research and multi-omics breeding approaches.

Why it matches plant phenotyping methodsソルガムの表現型データを標準化・統合し、可視化とアクセスを提供する研究基盤であり、表現型情報のデータ基盤として中心的です。

abstractSorghumBase (https://www.sorghumbase.org) is a knowledgebase designed for the sorghum research community.

SPROUT: A Scalable Diffusion Foundation Model for Agricultural Vision →

Image-based plant phenotyping depends on dense structural understanding of crops, yet pixel-level annotation remains expensive across species, organs, growth stages, and field conditions. General-purpose vision foundation models offer a natural route to label efficiency, but their web-scale pretraining objectives transfer weakly to agricultural imagery, where semantics are often determined by fine organ geometry inside repetitive, texture-dominated scenes. We introduce SPROUT, a diffusion foundation model for multi-crop plant phenotyping. SPROUT learns from 2.6 million unlabeled open-field images (MCD-2.6M) using a pixel-space Diffusion Transformer, and selects transferable features with a label-free effective-rank criterion over denoising timesteps. This design shifts pretraining from crop-based invariance to structure-preserving denoising, making the representation better aligned with dense phenotyping tasks. We evaluate SPROUT across dense phenotyping tasks, including organ segmentation, crop-weed parsing, depth estimation, and counting. SPROUT consistently improves over strong web-pretrained baselines, with the largest gains on dense structural prediction, and shows favorable label and compute efficiency compared with general-purpose and crop-specific foundation models. The source code and MCD-2.6M dataset are publicly available.

Why it matches plant phenotyping methods植物フェノタイピング向けの拡散基盤モデルを開発し、複数の作物・器官に対する画像ベースの構造推定タスクで評価しているため、表現学習法とデータセットが中心的な方法論的貢献である。

abstractWe introduce SPROUT, a diffusion foundation model for multi-crop plant phenotyping.

Dataset for orange fruit detection from UAV in citrus orchards. →

Accurate fruit detection in citrus orchards is essential for yield estimation, precision harvesting, and automated orchard monitoring. Although UAV-based imaging has become a powerful tool in precision agriculture, publicly available datasets for orange fruit detection remain scarce, particularly those integrating multispectral data under real field conditions. This lack of open resources limits the development and benchmarking of robust deep-learning models for cross-spectral and illumination-invariant detection. We present CampanetaOrangeFruit, a dataset acquired with a DJI Mavic 3 Multispectral UAV flying at 14 m above ground level over a commercial citrus orchard in Corbera, Valencia, Spain. The dataset comprises 550 synchronized captures (RGB + four multispectral bands: R, G, RE, NIR) for a total of 2750 images and 301,232 annotated orange instances. Each image includes YOLOv5-format annotations generated through a homography-based reprojection process, ensuring geometric consistency across spectral modalities. CampanetaOrangeFruit uniquely provides pixel-aligned, cross-spectral UAV imagery with fine-grained fruit-level annotations, enabling research on fruit detection, yield estimation, and domain adaptation in real-world orchard environments. It represents a valuable benchmark for advancing deep-learning approaches in precision agriculture and sustainable citrus production.

Why it matches plant phenotyping methods柑橘果実を対象としたUAV画像データセットとアノテーションを提供し、果実検出・収量推定モデルの開発およびベンチマークを中心課題とするため、植物フェノタイピング用データセットとして採用する。

abstractpublicly available datasets for orange fruit detection remain scarce

An AI-Driven Dual-Spectral Vision-Language Sensing Framework for Intelligent Agricultural Phenotyping. →

Seed varietal purity and physiological viability are critical determinants of crop yield and quality. However, non-destructive assessment faces significant challenges in fine-grained variety discrimination and the perception of internal defects. This study proposes S3-Net, an AI-driven multimodal sensing framework that integrates vision–language alignment with dual-spectral sensor fusion for autonomous seed quality evaluation. We introduce a Knowledge–Vision Alignment (KVA) module that incorporates encyclopedic morphological descriptions to guide feature learning, significantly enhancing few-shot generalization. Complementarily, a Dual-Spectral Fusion (DSF) module combines high-resolution RGB textures with penetrative Short-Wave Infrared (SWIR) sensing to jointly characterize external and internal traits. Experimental results on a custom multimodal dataset of 6000 samples across 12 crop categories demonstrate that S3-Net achieves 96.9% accuracy for species identification and 95.8% for viability detection. Notably, S3-Net outperforms ResNet-50 by 40.3% in extreme 1-shot scenarios. With a stable inference throughput of 95 fps, the system meets the high-throughput demands of industrial-scale applications, providing a robust and efficient solution for intelligent agricultural phenotyping.

Why it matches plant phenotyping methods種子の生理的状態（viability）をRGB・SWIR融合で非破壊推定するセンシング／AI手法が研究の中心であり、データセットと性能評価も提示しているため。種識別のみなら対象外だが、viability検出は植物状態のフェノタイピングに該当する。

abstractThis study proposes S3-Net, an AI-driven multimodal sensing framework that integrates vision–language alignment with dual-spectral sensor fusion for autonomous seed quality evaluation.

Enhancing plant disease detection through multi-modal integration of visual and textual data. →

Plant diseases pose a significant threat to global agriculture, impacting crop yields and quality. Early and accurate detection is essential for effective health management but remains challenging due to visual similarity among diseases and complex field backgrounds. This study introduces AgriMM, a novel multi-modal detection framework that integrates visual images with expert-validated textual descriptions to improve diagnostic precision. The framework features three key innovations: a Hybrid Convolutional-Attention Collaborative Backbone (HCACB) to capture both fine-grained lesions and global context; a Context-enhanced Visual-Language Path Aggregation Network (CVL-PAN) for multi-scale feature fusion; and an Adaptive Region-Text Contrastive Learning (AR-TCL) module to enforce precise semantic alignment. We constructed a comprehensive dataset comprising 30,000 images and detailed symptom descriptions across five major crops (tomato, cucumber, pepper, eggplant, and squash). Experimental results demonstrate that AgriMM achieves a mean Average Precision (mAP) of 95.2%, significantly outperforming state-of-the-art unimodal baselines by 11.6%. These findings confirm that integrating linguistic semantic priors effectively resolves visual ambiguity, providing a robust tool for precision agriculture and sustainable crop protection.

Why it matches plant phenotyping methods植物病害の症状を画像から検出・診断するマルチモーダル手法を開発し、データセットと性能比較で検証しているため、植物フェノタイピング手法が中心である。

abstractThis study introduces AgriMM, a novel multi-modal detection framework that integrates visual images with expert-validated textual descriptions to improve diagnostic precision.

Classification of Plant Leaf Diseases Using Image Processing and Machine Learning on Real-World Images →

Plant diseases pose a serious threat to global food security by directly affecting agricultural production. Traditional expert observation-based diagnosis processes are time-consuming, subjective, and error-prone, making early and accurate diagnosis difficult. This has necessitated the development of image processing and artificial intelligence-based systems that can automatically recognize disease symptoms from leaf images. This study aims to automatically classify plant leaf diseases using the PlantDoc dataset, which consists of images collected under real-world conditions. First, various image processing steps, such as denoising, color space transformations, segmentation, and contour detection, were applied to the leaf images to extract color, texture, and geometry-based features. The resulting features were classified using Support Vector Machines, Random Forests, and k-Nearest Neighbors, and the performance of these models was compared. Furthermore, a deep learning-based MobileNetV2 model was trained using transfer learning and data augmentation techniques and compared with classical methods. Experimental results show that the Random Forests model achieved the highest accuracy rate among classical methods, at 81.5%, while the MobileNetV2 model outperformed all other methods, with an accuracy rate of 86.9%. These findings demonstrate that deep learning-based approaches have higher generalization capabilities on complex, multi-class real-world data. Furthermore, classical methods, thanks to their interpretability and low computational cost, can be a good alternative in resource-limited systems.

Why it matches plant phenotyping methods植物葉の病徴を画像から抽出・分類する画像処理および機械学習手法が研究の中心であり、植物病害状態の表現型推定に該当する。

abstractThis study aims to automatically classify plant leaf diseases using the PlantDoc dataset, which consists of images collected under real-world conditions.

Simultaneous multi-disease detection from the same leaf: a generalized approach using deep learning and image splitting. →

Plant disease represents one of the most severe threats to agricultural production. Deep learning has emerged as a promising solution for automating the recognition of these diseases, leading to a richness of disease recognition applications based on deep learning. However, most existing applications do not address the challenge of simultaneous multi-disease detection from the same leaf. In this study, we introduce a deep learning-based model designed to detect and recognize multiple diseases from the same leaf simultaneously. Our method enables the recognition of each disease's symptoms separately from small leaf regions, independent of other diseases or specific crop types, through an isolation method. This approach also allows the model to generalize disease detection to new crops not encountered during training. Additionally, our method calculates the prevalence rate of each disease on the leaf and determines the overall extent of all diseases present. To evaluate the effectiveness of our approach, we applied it to the widely recognized PlantVillage dataset, creating a new version for training and testing with three CNN models: Small Inception, MiniVGGNet, and LeNet5. The results demonstrate that the Small Inception architecture outperformed the other two CNNs in terms of classification performance. Despite some class imbalances in the new dataset, which were addressed through the use of class weights, this approach significantly enhanced the model's performance. Furthermore, while the proposed method demonstrates high performance in controlled environments, though its consistency under real field conditions still warrants deeper investigation. Overall, the findings underscore the effectiveness of our method and highlight its potential as an efficient solution applicable across diverse agricultural contexts.

Why it matches plant phenotyping methods葉画像から複数病害の症状を分離・認識し、葉上の病害有病率と全体的な病害程度を算出する画像ベース手法が研究の中心であるため、植物表現型計測法として含める。

abstractwe introduce a deep learning-based model designed to detect and recognize multiple diseases from the same leaf simultaneously.

HASPNet: a hierarchically attentive signal-preserving network for papaya leaf disease classification with explainable deep learning. →

The accuracy of papaya leaf disease classification is of highest priority in early-stage plant health surveillance and green farming. This paper presents HASPNet, a hierarchically attentive signal-preserving network specially designed for fine-grained papaya leaf disease classification from the newly proposed BDPapayaLeaf Dataset of 2,159 high-resolution images of five pathological classes. The network introduces a coordinated hierarchical attention framework; by integrating residual feature fusion with sequential SE and CBAM modules, HASPNet synchronizes multi-scale signal preservation with dual-stage recalibration, allowing the model to isolate subtle pathological signatures while maintaining global structural integrity. The architecture is additionally optimized using Swish activation, depthwise separable convolutions, and a cosine warm-up learning rate schedule to produce efficient gradient flow and convergence stability. Exhaustive ablation experiments validate the critical contribution of each architectural block, and the complete HASPNet obtains an accuracy of 93.87% (corresponding to a 6.13% error rate), an F1-score of 94%, and a reduced inference time of 21.33 ms, by a large margin surpassing top state-of-the-art backbones like MobileNetV2, DenseNet121, Inception-V3, Xception, and ResNet50 in both performance and computational efficiency. Additionally, activation function experiments validate Swish as the optimal non-linearity for this task. Interpretability is enhanced using Grad-CAM visualizations, which validate the model's attention on disease-specific leaf regions. Given the lack of existing benchmarks for the BDPapayaLeaf Dataset, HASPNet is evaluated against standard CNN backbones (MobileNetV2, ResNet50, etc.) to establish a performance-complexity baseline, justifying its selection for resource-constrained agricultural environments. The results validate the model's domain adaptability, and it is a strong candidate for real-world agricultural diagnostic systems and a valuable addition to vision-based plant pathology.

Why it matches plant phenotyping methodsパパイヤ葉の病徴画像から病害状態を分類する深層学習手法を開発し、アブレーション、既存CNNとの比較、Grad-CAMで技術検証しているため、植物フェノタイピング手法が中心です。

abstractThis paper presents HASPNet, a hierarchically attentive signal-preserving network specially designed for fine-grained papaya leaf disease classification from the newly proposed BDPapayaLeaf Dataset of 2,159 high-resolution images of five pathological classes.

A Three-Year Multimodal Holistic Dataset For Horticultural Tomato Cultivation. →

Existing tomato datasets often focus on short-term experiments or lack integrated environmental and agronomic data. We present Horti-M3-Tomato, a comprehensive three-year dataset collected in Northeast China's greenhouse, including high-resolution RGB images, environmental sensor data (recorded every 30 minutes), soil conditions, and detailed agronomic records such as yield data and management practices. Spanning three growing seasons (2023-2025), the dataset integrates temporal imaging, environmental monitoring, soil data, and manual phenotypic and yield records. Horti-M3-Tomato supports research on growth dynamics, genotype-environment interactions, and provides a benchmark for AI-based phenotyping and precision horticulture. The dataset is openly available for further research in controlled-environment agriculture.

Why it matches plant phenotyping methodsトマトの画像・環境センサーデータ・手動表現型記録を統合したデータセットであり、AIベースの表現型解析のベンチマークとして明示されているため、表現型データ基盤が中心です。

abstractincluding high-resolution RGB images, environmental sensor data (recorded every 30 minutes), soil conditions, and detailed agronomic records such as yield data and management practices.

Deep learning-based seed germination prediction using morphological traits and RGB images. →

Seed selection constitutes the initial and one of the most critical steps in agricultural productivity. The identification of high-quality seeds is a labor-intensive and costly process that requires considerable expertise. Within the scope of smart farming applications, this study proposes a deep learning–based model designed to automate the seed selection process by accurately predicting seed germination capacity from seed images. The proposed model determines whether a seed will germinate using RGB images and morphological traits automatically extracted from these images. The dataset used in this study comprises a total of 3,645 images belonging to three different seed types. For each seed type (okra, eggplant, and tomato), 405 seed images were acquired from three distinct imaging sources (digital microscope, camera, and scanner), labeled, and subsequently sown in seed trays. The germination status of each sown seed was systematically monitored and matched with its corresponding image data. The dataset was partitioned into 80% training and 20% testing subsets. Following 5-fold stratified cross-validation, the proposed model achieved an average weighted F1-score of 0.95 on the training set and 0.93 on the testing set for germination capacity prediction. The performance of the proposed model was further compared with widely used deep convolutional neural network architectures, including VGG19, ResNet50, and EfficientNetB5. Comparative results demonstrate that the proposed model provides competitive and robust performance for seed germination prediction. Overall, the findings indicate that the proposed approach can effectively be utilized for automated seed germination prediction. Future research should evaluate the generalizability of the model by conducting performance assessments on additional seed types.

Why it matches plant phenotyping methodsRGB画像から種子の形態形質を自動抽出し、発芽状態を予測する深層学習手法が研究の中心であるため、植物フェノタイピング手法として含める。

abstractthis study proposes a deep learning–based model designed to automate the seed selection process by accurately predicting seed germination capacity from seed images.

VIS-NIR-SWIR Hyperspectral Imaging and Advanced Machine and Deep Learning Algorithms for a Controlled Benchmark of Bean Seed Identification and Classification. →

Reliable seed accession identification underpins germplasm conservation, traceability and breeding; however, conventional assays remain destructive, labour-intensive and difficult to scale. Here, visible-near-infrared-shortwave infrared (VIS-NIR-SWIR) hyperspectral imaging (HSI; 449.54-2399.17 nm; 563 bands) was used to classify 32 grain-legume accessions ( n = 3200 seeds; 100 seeds per accession), comprising 30 common bean ( Phaseolus vulgaris L.) landraces plus two outgroup legumes ( Vigna angularis (Willd.) Ohwi & Ohashi and Cajanus cajan (L.) Huth). Each seed was represented by one ROI-averaged spectrum obtained from mean representative pixels within a standardised 10 × 10 pixel window at the centre of each seed. A fixed stratified 70:30 seed-level training:test partition was used, with 70 seeds per accession ( n = 2240) reserved for fully independent training and 30 seeds per accession ( n = 960) reserved as a fully independent test set. Principal component analysis (PCA) captured 97.42% of the spectral variance in the first three components (PC1 = 63.34%, PC2 = 23.78%, and PC3 = 10.31%). One-versus-rest wavelength association mapping revealed a maximum R 2 of 0.775 at 461.37 nm, and ReliefF concentrated the strongest reduced-band signal within 449.54-456.30 nm and 577.02-597.54 nm. In the original ReliefF-selected 16-band benchmark, the subspace discriminant reached 68.25% macro-F1 and 68.54% balanced accuracy; after edge-band trimming, the alternative 16-band configuration decreased to 60.67% and 60.94%, respectively. With respect to the full-spectrum sensitivity benchmark, linear discriminant analysis achieved 96.35% balanced accuracy, followed by linear SVM (94.17%). Deep learning trained directly on the full 563-band spectra reached 84.90% test accuracy, 84.47% macro-F1, 86.27% precision and 84.90% recall, with MLP_Wide outperforming the convolutional, recurrent and attention-based alternatives. Overall, under controlled laboratory conditions, this benchmark shows that accession discrimination is driven mainly by visible-domain contrasts in the most compact representations, whereas the full spectral context remains important for the most confusable accessions and for cautious future sensor design. The reduced-band findings should therefore be interpreted as exploratory guidance for sensor design rather than as a validated deployment-ready specification.

Why it matches plant phenotyping methods豆類種子の識別・分類を目的に、ハイパースペクトル画像取得、波長選択、機械学習・深層学習をベンチマークしており、種子形質の計測・抽出手法が中心である。

abstractOverall, under controlled laboratory conditions, this benchmark shows that accession discrimination is driven mainly by visible-domain contrasts

OzBarley: A genetic and phenotypic data resource capturing the Australian barley breeding history. →

OzBarley is a comprehensive genotype-to-phenotype resource to support research and enhance barley breeding by integrating genotypic and phenotypic data for gene discovery. This publicly available dataset comprises genotypic data from historical and modern elite barley cultivars of significance to Australian barley breeding. The phenotypic component includes high-throughput imaging and X-ray CT-based spike analysis, enabling trait measurements such as plant growth dynamics and seed morphology. Users can leverage genome-wide association studies (GWAS) and genomic selection to identify genetic variants associated with agronomically important traits in the OzBarley datasets, thereby accelerating targeted breeding strategies. The dataset is accessible for download under CC-BY 4.0 license and users are invited to contribute new data when using OzBarley plant material in their research. Through its FAIR-compliant design (Findable, Accessible, Interoperable, Reusable), OzBarley represents a resource to protect genotypes of historical relevance, explore the genetic architecture of adaptation to dryland environments, and to enhance knowledge of the resilience, yield, and quality of barley cultivars under diverse environmental conditions, contributing to global food security and agricultural sustainability.

Why it matches plant phenotyping methods高スループット画像およびX線CTによる形質取得を含む、再利用可能な遺伝型・表現型データ資源であり、植物フェノタイピング手法とデータセットが中心です。

abstractThe phenotypic component includes high-throughput imaging and X-ray CT-based spike analysis, enabling trait measurements such as plant growth dynamics and seed morphology.

FloraSyntropy-net: scalable deep learning with novel FloraSyntropy archive for large-scale plant disease diagnosis. →

Early diagnosis of plant diseases is critical for global food safety, yet most AI solutions lack the generalization required for real-world agricultural diversity. These models are typically constrained to specific species, failing to perform accurately across the broad spectrum of cultivated plants. To address this gap, we first introduce the FloraSyntropy Archive, a large-scale dataset of 178,922 images across 35 plant species, annotated with 97 distinct disease classes. We establish a benchmark by evaluating numerous existing models on this archive, revealing a significant performance gap. We then propose FloraSyntropy-Net, a novel federated learning framework (FL) that integrates a Memetic Algorithm (MAO) for optimal base model selection (DenseNet201), a novel Deep Block for enhanced feature representation, and a client-cloning strategy for scalable, privacy-preserving training. FloraSyntropy-Net achieves a state-of-the-art accuracy of 96.38% on the FloraSyntropy benchmark. Crucially, to validate its generalization capability, we test the model on the unrelated multiclass Pest dataset, where it demonstrates exceptional adaptability, achieving 99.84% accuracy. This work provides not only a valuable new resource but also a robust and highly generalizable framework that advances the field towards practical, large-scale agricultural AI applications.

Why it matches plant phenotyping methods植物病害画像の大規模データセットと診断モデルを開発・ベンチマークしており、植物の病害状態を画像から推定する方法が中心である。

abstractWe establish a benchmark by evaluating numerous existing models on this archive

YOLOv9s-multi: orientation-based fruit selection for robotic apple thinning →

Accurate detection and orientation estimation of immature apples are crucial for effective thinning decisions in robotic apple thinning. Existing research either relies on computationally expensive RGB-D approaches with 3D geometric fitting to estimate orientation and size or requires multiple separate models for thinning decision, limiting their real-time performance on robotic platforms. To address these issues, a multi-task model, YOLOv9s-Multi, is proposed. First, this model integrates segmentation and keypoint heads to perform instance segmentation of immature apples and detect their calyx keypoint positions. Second, the detection head employs an efficient and lightweight Depthwise Convolution module (DWConv) to reduce model parameters while accurately capturing spatial features across channels. Finally, the orientation is derived from the segmentation centroid and calyx keypoint, enabling pixel-based fruit selection for thinning decision-making. This model is evaluated on a self-developed dataset that divides immature apples based on developmental stage: Flower-Retained Stage (FR-Stage) and Fruit-Visible Stage (FV-Stage). Results show that instance segmentation and keypoint AP@0.5 for FV-Stage are 89.3% and 86.4%, respectively, while for FR-Stage they are 71.6% and 79.8%. The model further achieves prediction accuracies of 92.80% (FV-Stage) and 72.59% (FR-Stage) within an acceptable error of 30 °. The pixel-based fruit selection method achieves 74.00% and 70.31% selection accuracy on the test and an additional measurement dataset, respectively. Compared with the baseline YOLOv9s-seg, the number of parameters is reduced by 11.4%. In contrast to 3D fitting methods, our approach provides lower computational complexity, faster inference speed, and higher accuracy. These results demonstrate that the proposed model can efficiently estimate the orientation of immature apples and perform fruit selection in close-range scenes and complex lighting environments, which are challenging for depth cameras to handle. The code and datasets are publicly available on GitHub: https://github.com/DIANSLEE/YOLOv9s-Multi.

Why it matches plant phenotyping methods未熟リンゴのセグメンテーション、萼点検出、重心との関係から果実の向きという器官形質を推定する画像解析手法が研究の中心であり、精度評価とデータセット検証も行っているため。

abstracta multi-task model, YOLOv9s-Multi, is proposed.

Evaluation of soybean sprouting growth vigor based on ZnONPs. →

Introduction Nanoparticle-induced treatments can promote seed germination and improve germination potential under environmental stresses such as drought and salinity. This study aimed to investigate the effects of Zinc oxide nanoparticles (ZnONPs) on soybean seed germination and to develop a precise evaluation method. Methods We developed a full-time sequence crop growth vitality monitoring system. Using germination rate and root length as primary evaluation indicators, we conducted full-time sequence germination vitality monitoring experiments on soybean seeds treated with ZnONPs. A dataset was constructed from images documenting embryonic root growth. The developed detection model was used to evaluate image detection accuracy during germination. Germination index and embryonic root length were also calculated. Further tests were performed on seeds exposed to 600 mg/L ZnONPs dispersion, followed by treatment with different concentrations of NaCl and PEG6000 solutions. Results At a concentration of 600 mg/L ZnONPs dispersion, soybean seeds showed the highest germination rate (an increase of 28%) and the longest radicle length (an increase of 42%). Compared with deionized water, the 600 mg/L ZnONPs dispersion accelerated initial germination time, increased germination rate, and enhanced radicle length under low-concentration stress. Discussion The results indicate that, at certain concentrations, ZnONPs dispersion positively influences soybean seed germination under varying salinity and drought conditions. We examined morphological and physiological changes in ZnONPs-treated seeds under stress, establishing a preliminary foundation for evaluating crop and variety vitality. These findings provide new insights that may contribute to improving soybean germination under simulated stress conditions, serving as a preliminary theoretical reference for potential applications in arid and saline environments.

Why it matches plant phenotyping methods発芽中の画像から発芽率・幼根長を抽出する連続モニタリングシステムと検出モデルを開発し、精度評価とデータセット構築を行っており、表現型取得手法が中心である。

abstractto develop a precise evaluation method

UE5-Forest: A Photorealistic Synthetic Stereo Dataset for UAV Forestry Depth Estimation →

Dense ground-truth disparity maps are practically unobtainable in forestry environments, where thin overlapping branches and complex canopy geometry defeat conventional depth sensors -- a critical bottleneck for training supervised stereo matching networks for autonomous UAV-based pruning. We present UE5-Forest, a photorealistic synthetic stereo dataset built entirely in Unreal Engine 5 (UE5). One hundred and fifteen photogrammetry-scanned trees from the Quixel Megascans library are placed in virtual scenes and captured by a simulated stereo rig whose intrinsics -- 63 mm baseline, 2.8 mm focal length, 3.84 mm sensor width -- replicate the ZED Mini camera mounted on our drone. Orbiting each tree at up to 2 m across three elevation bands (horizontal, +45 degrees, -45 degrees) yields 5,520 rectified 1920 x 1080 stereo pairs with pixel-perfect disparity labels. We provide a statistical characterisation of the dataset -- covering disparity distributions, scene diversity, and visual fidelity -- and a qualitative comparison with real-world Canterbury Tree Branches imagery that confirms the photorealistic quality and geometric plausibility of the rendered data. The dataset will be publicly released to provide the community with a ready-to-use benchmark and training resource for stereo-based forestry depth estimation.

Why it matches plant phenotyping methods樹木の枝・樹冠形状を対象とするステレオ深度推定データセットを開発し、画素単位の視差ラベルと実画像との比較検証を提供しており、植物構造の取得方法が中心である。

abstractWe present UE5-Forest, a photorealistic synthetic stereo dataset built entirely in Unreal Engine 5 (UE5).

Machine Vision–Based Deep Learning for Automated Crop Disease Classification in Precision Agriculture Article →

Abstract Cherry is widely cultivated but remains challenging to harvest due to economic and ecological constraints, especially in developing countries such as Pakistan. Climate change, limited use of technology, and foliar diseases worsened by pesticide use further reduce productivity, particularly during fruiting. Conventional disease assessment depends on expert observation and grower experience, making it subjective and time-consuming. A comprehensive evaluation was conducted on the PlantCity dataset, which contains 5,714 high-density, full-color RGB images collected under challenging conditions and categorized into 5 classes. We compared three approaches: deep learning pre-trained, transfer learning, and a machine learning pipeline. Models were evaluated by accuracy, precision, recall, F1-score, Cohen’s Kappa, inference time, FLOPs, and throughput. Grad-CAM was used to improve interpretability. Transfer learning using DenseNet169 achieved the highest performance, with 99.80% accuracy, 99.80% precision, 99.80% recall, and a Cohen’s Kappa of 99.74%. These results were significantly higher than those obtained by other deep learning architectures and handcrafted baselines. Grad-CAM heatmaps confirmed that the models focused their attention on pathological areas. The proposed transfer-learning-based framework, particularly DenseNet169, demonstrates state-of-the-art diagnostic accuracy and features a modular structure. This design enables deployment on both high-performance servers and resource-constrained embedded devices, thereby facilitating early disease detection in precision agriculture.

Why it matches plant phenotyping methods植物画像から病害状態を分類する画像ベース表現型推定が研究の中心であり、複数モデルの比較・性能評価とGrad-CAMによる検証を行っているため。

abstractConventional disease assessment depends on expert observation and grower experience, making it subjective and time-consuming.

LeafInst - Unified Instance Segmentation Network for Fine-Grained Forestry Leaf Phenotype Analysis: A New UAV based Benchmark →

Intelligent forest tree breeding has advanced plant phenotyping, yet existing research largely focuses on large-leaf agricultural crops, with limited attention to fine-grained leaf analysis of sapling trees in open-field environments. Natural scenes introduce challenges including scale variation, illumination changes, and irregular leaf morphology. To address these issues, we collected UAV RGB imagery of field-grown saplings and constructed the Poplar-leaf dataset, containing 1,202 branches and 19,876 pixel-level annotated leaf instances. To our knowledge, this is the first instance segmentation dataset specifically designed for forestry leaves in open-field conditions. We propose LeafInst, a novel segmentation framework tailored for irregular and multi-scale leaf structures. The model integrates an Asymptotic Feature Pyramid Network (AFPN) for multi-scale perception, a Dynamic Asymmetric Spatial Perception (DASP) module for irregular shape modeling, and a dual-residual Dynamic Anomalous Regression Head (DARH) with Top-down Concatenation decoder Feature Fusion (TCFU) to improve detection and segmentation performance. On Poplar-leaf, LeafInst achieves 68.4 mAP, outperforming YOLOv11 by 7.1 percent and MaskDINO by 6.5 percent. On the public PhenoBench benchmark, it reaches 52.7 box mAP, exceeding MaskDINO by 3.4 percent. Additional experiments demonstrate strong generalization and practical utility for large-scale leaf phenotyping.

Why it matches plant phenotyping methods森林葉の個体分割と表現型解析のためのUAV画像データセットおよび新規セグメンテーション手法を開発・評価しており、植物表現型取得が中心である。

abstractwe collected UAV RGB imagery of field-grown saplings and constructed the Poplar-leaf dataset, containing 1,202 branches and 19,876 pixel-level annotated leaf instances.

Multi-Platform LiDAR Comparative Assessment for Above-Ground Biomass and Carbon Estimation in Mediterranean Woody Crops →

Reliable aboveground biomass (AGB) estimates for woody crops are required for carbon accounting and MRV; however, it remains unclear how LiDAR modality and sampling geometry influence plot-scale and tree-scale AGB predictions in intensively managed orchards. We benchmarked four LiDAR modalities across three Mediterranean woody-crop sites in Córdoba (Spain), IFAPA, Doña María, and Villaseca using open national airborne laser scanning (PNOA/ALS), Riegl ALS, unmanned laser scanning (ULS), and mobile laser scanning (MLS). The field inventory used 58 fixed-area plots (20×50 m; 0.1 ha) collected in December 2024-January 2025 (1,867 trees) and species-specific allometries based on D2r to derive tree and plot AGB; carbon was computed using wood carbon fractions (0.445 olive; 0.457 almond) and CO2e via IPCC conversion. Plot-level LiDAR metrics (e.g., mean height, p95, maximum height, and cover proxies) were extracted from normalized point clouds and modeled with Random Forest, XGBoost, and an ensemble under an 80/20 train-test split. Mean field AGB differed among sites (33.89, 30.94 and 12.76 Mg ha−1 for Villaseca, Doña María, and IFAPA). In the provided summaries, XGBoost achieved the lowest errors at IFAPA (RMSE = 0.400 Mg ha−1; R2 = 0.994) and Villaseca (RMSE = 0.872 Mg ha−1; R2 = 0.995), whereas PNOA was competitive at Doña María (RMSE = 0.725 Mg ha−1; R2 = 0.994). The results support cross-platform LiDAR for orchard AGB mapping and identify conditions under which open national LiDAR can enable scalable MRV. In addition, we evaluated TreeQSM-based quantitative structure models (QSMs) as an independent tree-level 3D reconstruction approach and examined their site-dependent agreement with field inventory estimates.

Why it matches plant phenotyping methods複数のLiDARモダリティと解析手法を比較・ベンチマークし、樹木およびプロットの地上部バイオマスを推定する技術評価が研究の中心であるため。

abstractWe benchmarked four LiDAR modalities across three Mediterranean woody-crop sites

Field-based hyperspectral characterization of wetland plant diversity and vitality in Burullus Lagoon (Nile Delta, Egypt) →

Burullus Lagoon, situated in the Nile Delta of Egypt, is a Ramsar-listed wetland of high ecological importance, particularly in relation to its floristic diversity. This study presents a field-based hyperspectral characterization of wetland vegetation with the objective of establishing a reference spectral library to support biodiversity assessment and environmental monitoring. Hyperspectral reflectance measurements were obtained for 41 plant species selected from a total of 63 floristically identified taxa, based on ecological dominance, spatial recurrence across sampling sites (≥3 stands), and suitability for reliable field spectral acquisition. Spectroscopic data were collected from 44 stands representing lagoon shores, islets, and open-water habitats using an ASD FieldSpec spectroradiometer covering the 350–2500 nm spectral range. A set of vegetation indices was applied to evaluate key biophysical and biochemical properties associated with plant vitality, water status, and biomass. The results indicate that the red and near-infrared regions provide the highest discriminatory capability among species, whereas the shortwave infrared region exhibits more limited discriminatory capability. Dominant taxa, including Phragmites australis and Atriplex halimus , displayed elevated near-infrared reflectance, consistent with differences in canopy structure and biochemical composition. Most species showed vegetation index responses broadly indicative of healthy physiological conditions, although interspecific variability suggests contrasting stress responses among taxa. Overall, the study demonstrates the applicability of field-based hyperspectral data for species-level discrimination in wetland environments and delivers a curated spectral library to support biodiversity conservation and long-term ecosystem management at Burullus Lagoon.

Why it matches plant phenotyping methods野外ハイパースペクトル計測とスペクトルライブラリ構築が研究の中心で、植物の活力、水分状態、バイオマスなどの状態推定に用いているため、単なる生態調査ではなく植物表現型計測への実質的応用に該当する。

abstractThis study presents a field-based hyperspectral characterization of wetland vegetation with the objective of establishing a reference spectral library to support biodiversity assessment and environmental monitoring.

PlantSegNeRF: A few-shot, cross-species method for plant 3D instance point cloud reconstruction via joint-channel NeRF with multi-view image instance matching →

Organ segmentation of plant point clouds is a prerequisite for the high-resolution and accurate extraction of organ-level phenotypic traits. Although the fast development of deep learning has boosted much research on segmentation of plant point clouds, the existing techniques for organ segmentation still face limitations in resolution, segmentation accuracy, and generalizability across various plant species. In this study, we proposed a novel approach called plant segmentation neural radiance fields (PlantSegNeRF), aiming to directly generate high-precision instance point clouds from multi-view RGB image sequences for a wide range of plant species. PlantSegNeRF performed two-dimensional (2D) instance segmentation on the multi-view images to generate instance masks for each organ with a corresponding instance identification (ID). The multi-view instance IDs corresponding to the same plant organ were then matched and refined using a specially designed instance matching (IM) module. The instance NeRF was developed to render an implicit scene containing color, density, semantic and instance information, which was ultimately converted into high-precision plant instance point clouds based on volume density. The results proved that in semantic segmentation of point clouds, PlantSegNeRF outperformed the commonly used methods, demonstrating an average improvement of 16.1 %, 18.3 %, 17.8 %, and 24.2 % in precision, recall, F1-score, and intersection over union (IoU) compared to the second-best results on structurally complex datasets. More importantly, PlantSegNeRF exhibited significant advantages in instance segmentation. Across all plant datasets, it achieved average improvements of 11.7 %, 38.2 %, 32.2 % and 25.3 % in mean precision (mPrec), mean recall (mRec), mean coverage (mCov), and mean weighted coverage (mWCov), respectively. Furthermore, PlantSegNeRF demonstrates superior few-shot, cross-species performance, requiring only multi-view images of few plants to train models applicable to specific or similar varieties. This study extends organ-level plant phenotyping and provides a high-throughput way to supply high-quality 3D data for developing large-scale artificial intelligence (AI) models in plant science. • A comprehensive dataset of well-labeled two-dimensional (2D) images and point clouds dataset of plants was established, including various varieties and growth stages. 50 plant samples were collected for each type. • A novel multi-view image instance matching (IM) module was proposed to align plant organ instance identifications (IDs) across different viewpoints, serving as the foundation for organ-level instance segmentation. • A multi-channel instance neural radiance fields (NeRF) module with encoding color, semantic, and instance information was developed to achieve high-precision mapping of 2D image colors, semantics, and aligned instances into 3D space, enabling point cloud background removal and fine-grained segmentation of plant organs.

Why it matches plant phenotyping methods植物器官の3D点群再構成・インスタンス分割を開発し、セグメンテーション性能を検証する手法研究であり、器官レベル表現型抽出を直接支援するため。

abstractwe proposed a novel approach called plant segmentation neural radiance fields (PlantSegNeRF), aiming to directly generate high-precision instance point clouds from multi-view RGB image sequences for a wide range of plant species.

GLiMPSe: A low-cost, high-throughput and accurate field phenotyping system for maize architectural traits →

Maize phenotyping remains a major bottleneck in genetic analysis and breeding. Despite advances in drones, field robots, and gantry phenotyping systems, ultra-affordable, high-throughput, field-based maize phenotyping at single-plant resolution is still lacking, largely due to the high cost, complex deployment, and limited flexibility of existing platforms under heterogeneous field conditions. To address these challenges, we propose a novel paradigm that integrates DIY imaging devices with customized computer vision–based analytics, and present GLiMPSe ( G iraffe + Li zard M aize P henotyping S yst e m), including two end-to-end phenotyping modules for maize plant architecture (the Giraffe module) and leaf traits (the Lizard module). The imaging device in the Giraffe module are built from modular electronics and 3D-printed parts from local retailers to achieve high-quality image acquisition. The Giraffe and the Lizard modules operate at speeds of 15 seconds and 8 seconds per sample, with costs of $379.1 and $241.1, respectively. Both modules feature fine-tuned YOLOv11x segmentation models for reliable and robust target segmentation, followed by customized Python-based analytical pipelines that enable precise extraction and quantification of phenotypic traits. This methodology achieves high accuracies ( R² ) for five key traits, including plant height (0.928), heights of above-ear leaves (0.87∼0.958), ear height (0.925), above-ear leaf number (0.837), and leaf width (0.937). To enhance accessibility, we developed user-friendly graphical interfaces and publicly released manually annotated datasets and source code to support broader adoption and further innovation. This work provides a practical and accessible solution for high-throughput field phenotyping and offers new opportunities for democratizing crop phenomics through affordable, open-source technologies.

Why it matches plant phenotyping methods低コストな撮像装置、コンピュータビジョン解析、形質抽出パイプライン、GUI、データセットとコードを統合したトウモロコシ表現型測定システムの開発・検証が研究の中心である。

abstractwe propose a novel paradigm that integrates DIY imaging devices with customized computer vision–based analytics, and present GLiMPSe

A parametric maize leaf model quantifying morphology and geometry via 3D triangle mesh →

Maize leaf morphology is poorly investigated because quantifying maize leaf geometry is still an open question due to the complexity of the 3D curved shape. By utilization of geometric curves, maize leaf morphology can be effectively described parametrically and quantitatively. We divided maize leaf into three components: midrib, cross-section and blade contour. Each component is represented by parametric curves and controlled by a group of parameters. A 3D maize leaf model is generated by translation, rotation and scaling of the three components. We demonstrated the parametric maize leaf model allows the applications of leaf geometry analysis, leaf-level radiation capture simulation and dataset synthesis for phenotyping pipeline. The parametric maize leaf model is configurable, extensible and scalable, allowing it to be used in agricultural digital-twin and high-accuracy phenotyping. It also has potential to serve as a platform for maize biophysical and biomechanical studies. The code for 3D maize leaf model generation is available at https://github.com/xzcppm/parametric_maize_leaf.

Why it matches plant phenotyping methodsトウモロコシ葉の3D形態・幾何をパラメトリックにモデル化し、表現型解析用データ合成にも利用できる手法を開発しているため、植物表現型取得・解析手法が中心である。

abstractA 3D maize leaf model is generated by translation, rotation and scaling of the three components.

A stem-leaf segmentation method of maize plant point cloud based on region growing and leaf phenotypic parameters measurement →

Maize leaf phenotypic parameters effectively reflect the photosynthesis and growth information of maize plants, which is crucial for breeding superior maize varieties. Current challenges include separating stems and leaves from a single maize plant and accurately measuring the phenotypic parameters of maize leaves. This study proposes a stem-leaf segmentation method based on region growing, incorporating adaptive cuboid region growing and slice region growing, alongside techniques for measuring phenotypic parameters of maize leaves. First, terrestrial laser scanning (TLS) was employed to obtain three-dimensional (3D) point cloud data of maize at the five-leaf (V5) and six-leaf (V6) stages. The point cloud data were then preprocessed to isolate single plant point clouds. Next, the maize point clouds were pre-segmented into three categories-central point clouds, partially expanded leaf point clouds, and unexpanded leaf point clouds-using center-edge segmentation, statistical filtering, and leaf classification. Adaptive cuboid region growing was applied to segment the unexpanded leaf point clouds, while slice region growing was used for partially expanded leaves, with Euclidean clustering optimizing the leaf point clouds, completing the segmentation process. Finally, various methods-including clustering counting, point-to-point distance accumulation, point-to-line distance, vector angle, point cloud triangulation, and triangle area accumulation-were utilized to automatically measure the number of maize leaves, leaf length, leaf width, leaf inclination angle, and leaf area. Compared with other point cloud stem-leaf segmentation methods based on geometric features and common 3D point cloud deep learning models (PointNet++, PointTransformer), the method proposed in this paper performs better. The segmentation results indicated that the Precision (P), Recall (R) and F₁-Score (F₁) for stem-leaf segmentation of all maize plants at the V5 stage exceeded 92.00%, with average values of 96.87%, 97.08%, and 96.97%, respectively. At the V6 stage, P, R, and F₁ exceeded 95.00%, with averages of 97.73%, 97.01%, and 97.67%, respectively. The algorithm accurately measured the number of leaves at the V5 stage, while a small error was noted at the V6 stage, yielding a percentage error (PE) of 0.93%. Measurement accuracy for leaf length, width, and area at both growth stages was greater than 93.80%, 92.80%, and 89.50%, respectively. Measurement accuracy for leaf inclination angle was lower, at 82.00% and 88.02% for the V5 and V6 stages, respectively. The proposed methods for stem-leaf segmentation and measurement of leaf phenotypic parameters are fast and accurate, providing technical support for high-quality breeding and intelligent management of maize. Our point cloud data of maize and source code is available from https://github.com/lmj-cau/stem-leaf-segmentation.git.

Why it matches plant phenotyping methodsトウモロコシの3D点群から茎葉を分割し、葉数・長さ・幅・面積・傾斜角を自動推定する手法を開発・比較検証しており、植物表現型取得が研究の中心である。

abstractThis study proposes a stem-leaf segmentation method based on region growing, incorporating adaptive cuboid region growing and slice region growing, alongside techniques for measuring phenotypic parameters of maize leaves.

Development of an enhanced hybrid attention YOLOv8s small object detection method for phenotypic analysis of root nodules →

Nodule formation and their involvement in biological nitrogen fixation are critical features of leguminous plants, with phenotypic characteristics closely linked to plant growth and nitrogen fixation efficiency. However, the phenotypic analysis of root nodules remains technically challenging due to their small size, weak texture, dense clustering, and occlusion. To address these challenges, this study constructed a scanner-based imaging platform and optimized data acquisition conditions for high-resolution, high-consistency root nodule images under field conditions. In addition, A hybrid small-object detection method, SCO-YOLOv8s, was proposed, integrating Swin Transformer and CBAM attention mechanisms into the YOLOv8s framework to enhance global and local feature representation. Furthermore, an Otsu segmentation-based post-processing module was incorporated to validate and refine detection results based on geometric features, boundary sharpness, and image entropy, effectively reducing false positives and enhancing robustness in complex scenes. Using this integrated approach, over 3375 nodules were identified from a single plant sample in under 1 min, with extracted phenotypic features such as diameter, color, and texture. A total of 10,879 high-quality annotated images were collected from 39 peanut varieties across 14 provinces and 31 soybean varieties across 12 provinces in China, addressing the current lack of large-scale datasets for legume root nodules. The SCO-YOLOv8s model achieved a precision of 97.29 %, a mAP of 98.23 %, and an overall identification accuracy of 95.83 %. This integrated approach provides a practical and scalable solution for high-throughput nodule phenotyping, and may contribute to a deeper understanding of nitrogen fixation mechanisms.

Why it matches plant phenotyping methods根粒の画像取得・検出・セグメンテーション・形質抽出を統合した高スループット表現型解析手法を開発し、精度評価と大規模データセット構築も行っているため、方法が研究の中心である。

abstractthis study constructed a scanner-based imaging platform and optimized data acquisition conditions for high-resolution, high-consistency root nodule images under field conditions.

Improved YOLOv8 for multi-colored apple fruit instance segmentation and 3D localization →

Robotic apple harvesting requires precise instance segmentation and 3D localization, especially for multi-colored apples under complex orchard conditions with occlusions and variable lighting. Current deep learning methods lack robustness and accuracy for such scenarios, limiting automation. This study proposes improved YOLOv8-based models to enhance segmentation of multi-colored apples, combined with a high-precision 3D localization pipeline to advance practical robotic harvesting. To address these issues, this study collected apple images in three colors from two locations, creating a dataset of 5171 images. Four enhanced YOLOv8-based models—RA-YOLO, GA-YOLO, YA-YOLO, and MCA-YOLO—were proposed for segmenting red, green, yellow, and mixed multi-colored apples. RA-YOLO integrates the GD mechanism and EMBConv structure based on EfficientNet's MBConv. GA-YOLO replaces standard convolutions with dynamic serpentine convolution and adds the P6 layer for large object detection. YA-YOLO utilizes deformable convolution (DCNv2) and introduces the new attention mechanism MPCA. MCA-YOLO combines the P6 layer, DCNv2, and EMBConv structure, merging the strengths of other models. RA-YOLO, GA-YOLO, and YA-YOLO achieved mAP values of 95.2 %, 96.4 %, and 95.4 %, respectively, for single-colored apple instance segmentation, surpassing baseline models and those in existing literature. MCA-YOLO achieved mAP values of 95.6 %, 96.6 %, and 94.6 % for single-colored apples and 95.6 % for mixed multi-colored apples. Ablation experiments validated the necessity of each module. Finally, a high-precision 3D localization and shaping pipeline was developed, achieving an average localization error of 2.636 mm and a shaping error of 0.768 mm, enabling millimeter-level localization and sub-millimeter-level shaping for apple harvesting optimization.

Why it matches plant phenotyping methodsリンゴ果実のインスタンス分割、3D位置推定、形状推定を開発・検証しており、収穫対象の単なる検出を超えて果実形状という植物器官形質を定量化する手法が中心である。

abstractThis study proposes improved YOLOv8-based models to enhance segmentation of multi-colored apples, combined with a high-precision 3D localization pipeline

LKCAFormer: a lightweight transformer with large-kernel cooperative attention for the segmentation of field maize leaf diseases. →

In smart agriculture, accurate segmentation of maize-leaf diseases in real field imagery supports timely intervention, but remains challenging under cluttered backgrounds, uneven illumination, occlusion, and diverse lesion morphology. We present LKCAFormer, a lightweight encoder–decoder segmentation network that integrates two key components: (i) a three-stage Large-Kernel Cooperative Attention encoder (LK-COAT) that progressively enlarges the effective receptive field via large-kernel depthwise convolutions while preserving fine boundaries using cooperative channel–spatial gating; and (ii) a cross-scale decoder (CSDecoder) that fuses shallow edge/detail cues with deep semantics to refine lesion boundaries at low computational cost. We evaluate LKCAFormer on CD&S and a controlled single-leaf variant derived from it (Single-CD&S), using disease IoU as the primary endpoint. Robustness is further assessed on a 266-image complex-case subset curated from held-out test data, together with paired two-sided tests. On Single-CD&S, LKCAFormer achieves 76.23 ± 2.25 disease IoU and 86.70 ± 1.96 Dice, yielding a modest + 0.58 IoU gain over the strongest lightweight baseline (SwiftFormer). On the more challenging CD&S benchmark, LKCAFormer reaches 69.09 ± 1.65 disease IoU and 78.87 ± 2.13 Dice, outperforming the strongest baseline (SegFormer) by + 4.05 IoU; gains on the complex-case subset are statistically significant. LKCAFormer remains compact (3.68 M parameters; 1.13G FLOPs), corresponding to approximately 12.7% of U-Net’s parameters and 1.47% of its FLOPs, while retaining practical end-to-end throughput under a unified profiling protocol. Limitations include fixed dataset splits, the lack of cross-device latency/energy benchmarking, and the absence of multi-seed variability analysis. Future work will extend validation across crops and sensors and provide deployment-oriented, hardware-aware latency and energy evaluations.

Why it matches plant phenotyping methodsトウモロコシ葉の病斑・病害状態を画像から分割・定量化する手法を開発し、複数ベンチマークと複雑事例で性能検証しており、植物表現型取得が中心である。

abstractaccurate segmentation of maize-leaf diseases in real field imagery supports timely intervention

Enabling early identification of nutritional deficiencies in hazelnut orchards through a data-driven robotic framework →

Identifying nutritional deficiencies at an early stage is crucial for maximizing yield production and ensuring healthy plants. Conventional methods generally rely on time-consuming analysis conducted by agronomic experts. To address this challenge, this study presents a data-driven approach for the early identification of nutritional deficiencies in hazelnut orchards. Different custom datasets, composed of images acquired in a real hazelnut orchard as well as in a controlled laboratory environment, are collected, and the performance of five state-of-the-art machine learning models in early detecting nutritional deficiencies is compared. In particular, ResNet, DenseNet, MobileNet, EfficientNet, and ConvNext models, along with a baseline based on support vector machines, are considered. Data augmentation techniques are introduced to synthetically increase the datasets, and their effectiveness is extensively evaluated. Additionally, a pipeline is designed to carry out the early identification of nutritional deficiencies onboard an agricultural robot. Experimental results on the early identification show that ConvNext achieves the highest performance: 81.79% accuracy and 0.8168 F1 score on a real-world dataset with four classes, and 75.54% accuracy with 0.7552 F1 score for the more challenging six-class scenario. Furthermore, the effectiveness of the integrated system is validated in preliminary laboratory experiments using a Turtlebot2 mobile base and a Franka Research 3 arm, equipped with RGB-D cameras. • Data-driven pipeline detects hazelnut nutrient deficiencies from leaf images. • Real and lab-acquired hazelnut leaf image datasets collected and publicly released. • ConvNext achieves 85.50% accuracy on 4-class hazelnut deficiencies in lab conditions. • 75.54% accuracy on 6-class real orchard dataset validates field robustness. • Two-stage pipeline with leaf detection and classification enables onboard robot monitoring.

Why it matches plant phenotyping methods葉画像から植物の栄養欠乏状態を推定する画像解析・機械学習パイプラインと、ロボット搭載システム、データセットを中心的に開発・評価しているため。

abstractthis study presents a data-driven approach for the early identification of nutritional deficiencies in hazelnut orchards

DeepPhenoTree – Apple Edition: a Multi-site apple phenology RGB annotated dataset with deep learning baseline models →

Abstract In machine learning–driven plant phenotyping, well-annotated image datasets are essential for developing robust models capable of capturing phenological variability across environments. Here, we introduce DeepPhenoTree – Apple Edition, a multi-site, multi-variety RGB image dataset dedicated to the detection of key phenological stages in apple trees. The dataset comprises 48,320 time-stamped RGB images acquired across four European orchards of the Apple REFPOP consortium under contrasting climatic conditions. From this large corpus, a carefully curated subset of 808 representative images was manually annotated. It includes 241,600 expert annotations covering developmental stages from dormant bud to fruit maturity. Images were acquired using a standardized tractor-mounted phenotyping platform equipped with active flash illumination, ensuring consistent lighting conditions across sites and acquisition dates. Phenological structures were annotated following the BBCH scale, with bounding boxes adapted to organ visibility and developmental stage. In addition to the dataset, we provide baseline deep learning experiments to illustrate detection performance and assess model generalization across locations.

Why it matches plant phenotyping methodsリンゴの生育ステージを検出する注釈付き画像データセットと、標準化された撮像プラットフォームおよびベースラインモデルを提供しており、植物フェノタイピング手法・再利用可能データが研究の中心である。

abstractHere, we introduce DeepPhenoTree – Apple Edition, a multi-site, multi-variety RGB image dataset dedicated to the detection of key phenological stages in apple trees.

Transformer model to determine spatio-temporal relationships of variables, and interpretability for soybean seed yield, oil, and protein prediction. →

Accurate in-season prediction of seed yield and seed composition traits such as oil and protein are useful for gaining accuracy and efficiency in soybean breeding. These predictions can also inform farmers, enabling them to improve their field management practices, and guide their market decisions. We report a Transformer-based deep learning framework built on 30 years of multi-environment performance data from the Northern and Southern Uniform Soybean Tests (UST) across North America. Unlike earlier studies on seed yield, oil and protein prediction that focus on limited years, regions, single modalities, we utilized a comprehensive dataset that includes weather, genotype, and management factors, ensuring a more holistic approach to soybean yield, oil, and protein prediction. Our model integrates multivariate time-series weather data with genotypic relationship information, maturity group, and geographic location, to predict variety performance in diverse environments. Our model captures complex temporal patterns associated with trait variability; showing high predictive accuracy (R2) of 77.6 ± 0.2%, 63.9 ± 4.7%, and 79.3 ± 2.3% for seed yield, oil, and protein, respectively. Additionally, for seed yield, we also evaluated multiple interpretability methods to assess feature importance for predictor variables and critical growing timepoints, and solar radiation and temperature were noted as the key predictors. Overall, these results demonstrate the usefulness of a Transformer-based model in trait predictions, and the utility of large cooperative datasets from breeding programs.

Why it matches plant phenotyping methodsTransformerによるダイズ収量・油・タンパク質形質の予測フレームワークが研究の中心であり、植物形質の計算的推定手法を開発・評価している。

abstractWe report a Transformer-based deep learning framework built on 30 years of multi-environment performance data

Automatic tree-level based forest inventories retrieval via ultra-high resolution UAV images and deep learning →

Forest inventories play an essential role in managing and protecting forest resources as well as quantifying carbon stocks. Recent advances in Uncrewed Aerial Vehicles (UAVs) have enhanced capabilities for efficiently monitoring forest dynamics across large geographic areas. RGB cameras are typically preferred for rapid and scalable forest inventory missions owing to three distinct advantages, including low cost, ease of use, and high resolution. However, compared with multispectral or hyperspectral sensors, the limited spectral signals of RGB cameras pose challenges for tree crown detection and classification. The ability of deep learning methods to capture structural and contextual cues from imagery helps alleviate some of the limitations of RGB data. In this study, we propose an Individual Tree Crown (ITC)-based framework leveraging UAV data and advanced deep learning models for inventories of individual trees in dense and natural forests. First, we develop the ITC-based Multi-Task Convolutional Neural Network (ITCMNet), which incorporates multi-scale contexts to simultaneously and accurately identify individual tree crowns, discriminate tree species, and assess tree vitality. Second, structural parameters for each individual crown are extracted to estimate forest carbon storage using species-specific allometric models. Unlike conventional pixel-based methods, our proposed ITCMNet enables precise forest investigations at the ITC level, enhancing both performance and interpretability. We collected a comprehensive dataset consisting of 2456 ultra-high resolution (1.6 cm) UAV RGB images and 27,160 labeled trees across 105 plots distributed in three dense forests and one city park in Germany to evaluate our framework. The ITCMNet demonstrated robust tree crown delineation performance, achieving an F1 score of 0.81. Additionally, our method attained an F1 score in species classification (i.e., 0.54 for angiosperms and 0.76 for gymnosperms) and vitality identification (0.66). Utilizing precise tree parameters, species information, and species-specific allometric models, our carbon storage estimation surpassed current satellite-based carbon products. The carbon stock estimation achieved an R 2 of 0.83 and the carbon storage range in the Bamberg forests is approximately 50 to 110 Mg C/ha. These results show that our proposed framework provides detailed, cost-effective forest inventories, highlighting its potential to support various downstream forestry applications. The dataset and source code are available ( https://www.dlr.de/en/eoc/about-us/remote-sensing-technology-institute/photogrammetry-and-image-analysis/public-datasets/bamforests ; https://github.com/WendyFan52/ITCMNet ).

Why it matches plant phenotyping methodsUAV画像と深層学習により個体樹冠を検出し、樹冠構造、樹種、樹勢などの植物状態を抽出する枠組みを開発・評価しており、植物フェノタイピング手法が中心である。

abstractwe propose an Individual Tree Crown (ITC)-based framework leveraging UAV data and advanced deep learning models for inventories of individual trees in dense and natural forests.

ZamYOLO-maize: a YOLOv8n-based deep learning framework for automated detection and classification of maize leaf diseases in field conditions in Zambia. →

Maize, a critical staple crop in Zambia, faces persistent threats from foliar diseases such as Gray Leaf Spot, Northern Corn Leaf Blight, and Maize Streak Virus, significantly affecting smallholder productivity. Limited access to expert diagnostics, coupled with complex field conditions including occlusions and variable lighting, necessitates accessible, real-time disease detection systems tailored to local environments. To address this gap, this study first developed a novel field-captured dataset of Zambian maize leaf images, annotated with bounding boxes for disease lesions and labeled by disease type and severity to reflect real-world agri-ecological variability. Building on this dataset, we propose ZamYOLO-Maize, a multi-stage automated diagnostic framework integrating lesion detection, hierarchical disease classification, and severity assessment. A comparative evaluation was conducted using four state-of-the-art object detection models: YOLOv5n, YOLOv8s, YOLOv10s, and YOLOv8n, with performance assessed using precision, recall, F1-score, and inference speed. Experimental results demonstrate that YOLOv10s achieved the highest predictive performance (Precision = 0.997, Recall = 0.999, F1-score = 0.999), while YOLOv8n provided the optimal trade-off for edge deployment, achieving the fastest inference speed (4.65 ms/image) with a competitive F1-score of 0.995. The framework exhibited strong robustness under field variability, confirming its practical applicability. By integrating a locally representative dataset with an efficient deep learning pipeline, this study establishes a scalable foundation for mobile-based maize disease diagnostics, contributing to precision agriculture and supporting food security initiatives in Zambia and comparable agricultural regions.

Why it matches plant phenotyping methodsトウモロコシ葉画像から病斑、病害種、重症度を推定するデータセットと深層学習フレームワークを開発・比較評価しており、植物病害表現型の取得・抽出が中心です。

abstractthis study first developed a novel field-captured dataset of Zambian maize leaf images, annotated with bounding boxes for disease lesions and labeled by disease type and severity

BloomNet: Exploring Single vs. Multiple Object Annotation for Flower Recognition Using YOLO Variants →

Precise localization and recognition of flowers are crucial for advancing automated agriculture, particularly in plant phenotyping, crop estimation, and yield monitoring. This paper benchmarks several YOLO architectures such as YOLOv5s, YOLOv8n/s/m, and YOLOv12n for flower object detection under two annotation regimes: single-image single-bounding box (SISBB) and single-image multiple-bounding box (SIMBB). The FloralSix dataset, comprising 2,816 high-resolution photos of six different flower species, is also introduced. It is annotated for both dense (clustered) and sparse (isolated) scenarios. The models were evaluated using Precision, Recall, and Mean Average Precision (mAP) at IoU thresholds of 0.5 (mAP@0.5) and 0.5-0.95 (mAP@0.5:0.95). In SISBB, YOLOv8m (SGD) achieved the best results with Precision 0.956, Recall 0.951, mAP@0.5 0.978, and mAP@0.5:0.95 0.865, illustrating strong accuracy in detecting isolated flowers. With mAP@0.5 0.934 and mAP@0.5:0.95 0.752, YOLOv12n (SGD) outperformed the more complicated SIMBB scenario, proving robustness in dense, multi-object detection. Results show how annotation density, IoU thresholds, and model size interact: recall-optimized models perform better in crowded environments, whereas precision-oriented models perform best in sparse scenarios. In both cases, the Stochastic Gradient Descent (SGD) optimizer consistently performed better than alternatives. These density-sensitive sensors are helpful for non-destructive crop analysis, growth tracking, robotic pollination, and stress evaluation.

Why it matches plant phenotyping methods花の物体検出モデルを比較評価し、注釈付きFloralSixデータセットを導入することが中心で、花の認識・局在化を植物フェノタイピングへ応用する技術研究である。

abstractThis paper benchmarks several YOLO architectures such as YOLOv5s, YOLOv8n/s/m, and YOLOv12n for flower object detection under two annotation regimes: single-image single-bounding box (SISBB) and single-image multiple-bounding box (SIMBB).

RiceDetect-Net: a lightweight real-time detection framework for rice diseases. →

Rice disease detection is vital for food security, prevention efficiency, pesticide reduction, and sustainable agriculture. Challenges like poor model applicability, low accuracy, and limited datasets make this research essential. Existing models face issues with large parameters, complex computations, and insufficient semantic information capture. This paper introduces a large rice disease dataset and proposes the RiceDetect-Net model to address these challenges. The model integrates a brand-new lightweight detection head LE-Head to reduce the parameter quantity and computational complexity. To boost accuracy, the model integrates the newest FCA attention mechanism into its high-level semantic processing component, strengthening its capacity to interpret complex semantic data. Testing on a custom rice disease dataset comprising 54,240 images, the model attained 94.3% accuracy with a parameter count of 2.32 M. The enhanced model achieves a 0.4% increase in accuracy while reducing parameters by 10% relative to the baseline YOLOv11. The detection model is more lightweight, can adapt to the computing power of field detection equipment, is more suitable for practical scenario applications, and provides technical support for the development of smart agriculture.

Why it matches plant phenotyping methodsイネ病害を画像から検出するモデルと大規模データセットの開発・評価が中心で、植物の病害状態を推定する画像ベース表現型手法に該当する。

abstractThis paper introduces a large rice disease dataset and proposes the RiceDetect-Net model to address these challenges.

Comprehensive Review of Machine Learning and Deep Learning Methods for Plant Disease Detection via PlantVillage Dataset →

Plant diseases continue to pose a serious challenge to agriculture, leading to substantial yield losses and posing a threat to global food security. Accurate and early identification of plant diseases is crucial for effective crop management. However, traditional manual inspection methods are time-consuming, subjective, and heavily dependent on expert knowledge. With recent advances in artificial intelligence and computer vision, automated plant disease detection systems have emerged as a reliable alternative. These systems depend strongly on large, well-annotated image datasets for training and evaluation. Among publicly available resources, the PlantVillage dataset is one of the most widely used benchmarks for plant disease research. It contains over 50,000 high-resolution RGB leaf images captured under controlled conditions and covers 14 crop species, including tomato, potato, and bell pepper, along with multiple other plants. The dataset represents 38 distinct classes encompassing both healthy and diseased leaf categories. All images are collected against uniform backgrounds, providing visual consistency and making the dataset suitable for benchmarking machine learning and deep learning–based plant disease classification models. This survey provides a comprehensive review of the PlantVillage dataset and its contribution to the advancement of automated plant disease detection. It traces the progression from traditional handcrafted feature–based classifiers to convolutional neural networks, transfer learning approaches, and recent transformer-based architectures. Various methods are compared in terms of classification accuracy, generalization ability, computational efficiency, and robustness. In the survey we have identified that Transfer Learning model showed 99.75% accuracy. The survey also discusses key limitations of the dataset, particularly its controlled imaging conditions and challenges related to real-field deployment. Finally, future research directions are highlighted, including domain adaptation, explainable artificial intelligence, multi-disease recognition, and real-world agricultural applications. This work aims to offer researchers a structured understanding of the PlantVillage dataset and support the development of next-generation intelligent crop disease diagnostic systems.

Why it matches plant phenotyping methodsPlantVillage画像を用いた植物病害状態の画像ベース推定手法とデータセットを中心に、複数の機械学習手法を比較・レビューしているため、植物フェノタイピング方法論のレビューとして含める。

abstractThis survey provides a comprehensive review of the PlantVillage dataset and its contribution to the advancement of automated plant disease detection.

Early detection of strawberry and grape diseases under real-world field conditions using deep learning →

Abstract Crop diseases remain a significant threat to agricultural productivity and fruit quality, particularly for high-value crops such as strawberries and grapes. Early and reliable detection of these diseases under real-world conditions is essential but remains challenging due to variations in environment, illumination, and imaging perspectives. Leveraging recent advances in deep learning and computer vision, this study presents a robust framework for the automated detection and classification of strawberry and grape diseases using convolutional neural network (CNN) models, i.e., VGG16, ResNet101v2, InceptionV3, and DenseNet121. Unlike many existing works that rely solely on controlled or publicly available datasets, we constructed two specialized datasets by combining field-captured images under diverse environmental conditions with online sources, thereby enhancing robustness and ecological validity. The strawberry dataset includes six disease classes, while the grape dataset encompasses seven classes, covering economically significant pathologies such as anthracnose, black rot, gray mold, powdery mildew, sour rot, and leaf scorch. Extensive experiments were conducted using state-of-the-art CNN architectures, including VGG16, ResNet101v2, InceptionV3, and DenseNet121. On strawberries, DenseNet121 and InceptionV3 achieved accuracies of 94% (training) and 95% (testing), respectively, while VGG16 delivered superior performance on grapes, achieving 95% (training) and 92% (testing). Beyond technical accuracy, the proposed models were explicitly designed for applicability in actual field conditions, ensuring that the system can be directly adapted for use by farmers and plant pathologists as a practical decision-support tool. The findings provide a foundation for scalable, automated, and field-ready disease detection systems, contributing to more sustainable, data-driven crop management practices.

Why it matches plant phenotyping methods植物画像からイチゴ・ブドウ病害を自動検出・分類する深層学習手法を開発し、実圃場画像を含むデータセットで性能評価しているため、植物病害表現型の取得・推定が中心である。

abstractthis study presents a robust framework for the automated detection and classification of strawberry and grape diseases using convolutional neural network (CNN) models

Comprehensive phenomics and vegetative yield analysis of global kale (Brassica oleracea var. acephala) germplasm in controlled environment agriculture. →

BACKGROUND: Kale (Brassica oleracea var. acephala) is a high value leafy vegetable with an extensive domestication history and germplasm diversity, making it an ideal target for genetic improvement. To meet growing food security needs particularly with controlled environment agriculture (CEA) systems, specialized breeding strategies are required. The goal of this study was to survey the phenotypic architecture of a global kale germplasm collection under commercial CEA conditions. This study establishes a phenotypic baseline and serves as a hypothesis generating resource for future genetic and physiological studies in kale and other leafy vegetables grown under CEA. RESULTS: A total of 203 kale accessions were phenotyped for 113 quantitative traits using high-throughput phenotyping methods. Significant differentiation was observed across all traits, with coefficient of variation ranging from 2.5% to 180.7%, confirming broad genetic variability among accessions. Trait correlation networks and hierarchical clustering grouped phenotypes into seven biologically corresponding modules including leaf, stem and root morphology, plant architecture, hyperspectral indices, and seedling growth. These modules highlight coordinated phenotypic patterns among traits. Integrative yield analyses combining partial least squares variable importance in projection with differential trait analysis identified 28 phenotypes most strongly associated with total aboveground fresh weight, a robust proxy for CEA vegetative yield. Principal component analysis further distilled these traits into three orthogonal components explaining 87.1% of total yield variation. These components represented modules related to plant organ size, canopy structure, and density, emphasizing their biological contribution to harvestable biomass. CONCLUSIONS: This study generates a foundational phenomics resource and comprehensive dissection of kale’s yield architecture under CEA conditions. The composition of traits identified constitutes a targeted set of breeding traits to be further validated for improved leafy vegetable yield. By integrating large-scale germplasm resources with phenomics, this work establishes the utility of a high-throughput phenotypic analysis for further leafy crop research and improvement.

Why it matches plant phenotyping methods大規模なハイスループット植物表現型解析を中核とし、113形質の取得、統合解析、再利用可能なフェノミクス資源の構築を行っているため。

abstractA total of 203 kale accessions were phenotyped for 113 quantitative traits using high-throughput phenotyping methods.

Grow with the Flow: 4D Reconstruction of Growing Plants with Gaussian Flow Fields →

Modeling the time-varying 3D appearance of plants during growth poses unique challenges: unlike most dynamic scenes, plants continuously generate new geometry as they expand, branch, and differentiate. Existing dynamic scene representations are ill-suited to this setting: deformation fields provide insufficient constraints to yield physically plausible scene dynamics, and 4D Gaussian splatting represents the same physical structures with different Gaussian primitives at different times, breaking temporal consistency. We introduce GrowFlow, a dynamic representation that couples 3D Gaussian primitives with a neural ordinary differential equation to model plant growth as a continuous flow field over geometric parameters (position, scale, and orientation). Our representation enables consistent appearance rendering and models nonlinear, continuous-time growth dynamics with full temporal correspondences for every primitive. To initialize a sufficient set of Gaussian primitives, we first reconstruct the mature plant and then learn a reverse-growth process, effectively simulating the plant's developmental history in reverse. GrowFlow achieves superior image quality and geometric coherence compared to prior methods on a new, multi-view timelapse dataset of plant growth, and provides the first temporally coherent representation for appearance modeling of growing 3D structures.

Why it matches plant phenotyping methods植物の成長を対象に、4D再構成と連続的な成長表現を開発し、幾何学的整合性と画像品質を既存手法・新規データセットで比較評価しているため、植物フェノタイピング手法が中心です。

abstractWe introduce GrowFlow, a dynamic representation that couples 3D Gaussian primitives with a neural ordinary differential equation to model plant growth as a continuous flow field over geometric parameters (position, scale, and orientation).

NucVerse3D: Generalizable 3D nuclear instance segmentation across heterogeneous microscopy modalities →

Abstract Accurate three-dimensional (3D) nuclear instance segmentation is a prerequisite for quantitative phenotyping in volumetric microscopy, yet remains challenging in densely packed tissues, irregular nuclear morphologies, and across heterogeneous imaging modalities. Here we present NucVerse3D, a deep-learning framework for generalized 3D nuclei instance segmentation that combines a residual attention 3D U-Net architecture with a reversible gradient-field representation for robust centroid-aware instance reconstruction. NucVerse3D is trained end to end in 3D using modality-agnostic preprocessing and isotropic scale normalization, enabling deployment across confocal microscopy, two-photon microscopy, light-sheet microscopy, micro–computed tomography, and scanning electron microscopy volumes. We benchmarked NucVerse3D on seven volumetric datasets spanning multiple species and tissues, comprising more than forty thousand manually annotated nuclei, including newly released ground-truth datasets of mouse liver tissue (control and hepatocellular carcinoma) and Drosophila brain glial nuclei. Across datasets, NucVerse3D achieved consistently high precision, recall, F1-score, and average precision, and outperformed the state-of-the-art methods particularly in dense and irregular settings, while remaining competitive on simpler cases. A single generalized model trained on pooled data matched the performance of dataset-specific models, and ablation experiments demonstrated that preprocessing and scale normalization substantially contribute to performance under strict intersection-over-union criteria. To demonstrate the biomedical utility of NucVerse3D, we applied it to three-dimensional liver images from a mouse model of hepatocellular carcinoma (HCC). High-fidelity, nucleus-by-nucleus segmentation enabled the quantification of the Nuclear Decoupling Score (NDS), which captures deviations in nuclear DNA–volume coupling at the single-nucleus level. NDS analysis revealed a progressive increase in nuclear abnormalities within tumor regions, forming spatially coherent domains of dysregulated nuclei and highlighting NDS as a potential quantitative biomarker of dysplastic and tumor tissue. Together, NucVerse3D provides a robust and generalizable solution for 3D nuclear instance segmentation and enables quantitative nuclear phenotyping across imaging modalities. Highlights - NucVerse3D provides accurate 3D nuclear instance segmentation across modalities - Residual attention and gradient fields enable robust separation of dense nuclei - New 3D annotated datasets of mouse liver and Drosophila brain are released - A generalized model achieves performance comparable to dataset-specific training - 3D nuclear phenotyping reveals spatially organized nuclear abnormalities in HCC

Why it matches plant phenotyping methods3D核インスタンスセグメンテーション手法を開発し、多数のデータセットでベンチマークするとともに、核形態状態の定量的フェノタイピングへ応用しているため。

abstractHere we present NucVerse3D, a deep-learning framework for generalized 3D nuclei instance segmentation

Transferability and Robustness in Proximal and UAV Crop Imaging →

AI-driven imaging is becoming central to crop monitoring, with proximal and unmanned aerial vehicle (UAV) platforms now routinely used for disease and stress detection, yield estimation, canopy structure, and fruit counting. Yet, as these models move from plots to farms, the main bottleneck is no longer raw accuracy but robustness under distribution shift. Systems trained in one field, season, cultivar, or sensor often fail when the scene, sensor, protocol, or timing changes in realistic ways. This review synthesizes recent advances on robustness and transferability in proximal and UAV imaging, drawing on a corpus of 42 core studies across field crops, orchards, greenhouse environments, and multi-platform phenotyping. Shift types are organized into four axes, namely scene, sensor, protocol, and time. The article also maps the empirical evidence on when RGB imaging alone is sufficient and when multispectral, hyperspectral, or thermal modalities can potentially improve robustness. This serves as a basis to synthesize acquisition and evaluation practices that often matter more than architectural tweaks, which include phenology-aware flight planning, radiometric standardization, metadata logging, and leave-one-field/season-out splits. Adaptation options are consolidated into a practical symptom/remedy roadmap, ranging from lightweight normalization and small target-set fine-tuning to feature alignment, unsupervised domain adaptation, style translation, and test-time updates. Finally, a benchmark and dataset agenda are outlined with emphasis on object-oriented splits, cross-sensor and cross-scale collections, and longitudinal datasets where the same fields are followed across seasons under different management regimes. The goal is to outline practices and evaluation protocols that support progress toward deployable and auditable systems, noting that such claims require standardized out-of-distribution testing and transparent reporting as emphasized in the benchmark specification and experiment suite proposed here.

Why it matches plant phenotyping methods植物の近接・UAV画像による病害・ストレス・収量・キャノピー構造・果実数の推定について、頑健性、転移性、取得・評価プロトコル、ベンチマークを体系化する方法論レビューであり、フェノタイピング手法が中心です。

abstractThis review synthesizes recent advances on robustness and transferability in proximal and UAV imaging, drawing on a corpus of 42 core studies across field crops, orchards, greenhouse environments, and multi-platform phenotyping.

Automatic pixel-level annotation for plant disease severity estimation →

Plant disease adversely impacts food production and quality. Alongside detecting the disease, estimating its severity is important in managing the disease. Artificial intelligence deep learning-based techniques for plant disease detection are emerging. Unlike most of these techniques, which focus on disease recognition, this study addresses various plant disease-related tasks, including annotation, severity classification, lesion detection, and leaf segmentation. We propose a novel approach that learns the disease symptoms, which are then used to segment disease lesions for severity estimation. To demonstrate the work, a dataset of barley images was used. We captured the images of barley plants inoculated with diseases on test-bed paddocks at various growth stages. The dataset was automatically annotated at a pixel level using a trained vision transformer to obtain the ground truth labels. The annotated dataset was applied to train salient object detection (SOD) methods. Two top-performing lightweight SOD models were used to segment the disease lesion areas. To evaluate the performance of the SODs, we have tested them on our dataset and several other datasets, including the Coffee dataset, which has expert pixel-level labels that were unseen during the training step. Several morphological and spectral disease symptoms, including those akin to the widely used ABCD rule for human skin-cancer detection, i.e., asymmetry (A), border irregularity (B), colour variance (C), and diameter (D), are learned. To the best of our knowledge, this is the first study to incorporate these ABCD features in plant disease detection. We further extract visual and texture features using the grey level co-occurrence matrix (GLCM) and fuse them with the ABCD features. For the coffee dataset, our method achieved 82 + % detection accuracy on the severity classification task. The results demonstrate the performance of the proposed method in detecting plant diseases and estimating their severity.

Why it matches plant phenotyping methods植物病斑の画素レベル分割と病害重症度推定という、植物状態を画像から定量化する手法が研究の中心であり、データセット構築・モデル評価も行っている。

abstractthis study addresses various plant disease-related tasks, including annotation, severity classification, lesion detection, and leaf segmentation

Light-resilient visual regression of strawberry ripeness for robotic harvesting →

Accurate real-time quantification of strawberry ripeness is critical for advancing selective strawberry harvesting robots. However, existing methods often overlook inconsistencies caused by color distortion under varying light intensities. This study aims to verify and quantitatively analyze the influence of illumination on strawberry ripeness, and further presents a novel light-resilient vision-based ripeness regression method that overcomes these limitations through three key innovations. First, the research established the first fine-grained ripeness metric under controlled lighting conditions and introduced the comprehensive LightStrawberry dataset, featuring multi-illumination strawberry images. Second, we propose SRR-Net, an innovative end-to-end strawberry ripeness regression network built upon YOLOv8/YOLOv11. The network incorporates a dedicated ripeness regression branch that operates in parallel with the detection and segmentation heads, enabling simultaneous and efficient estimation of strawberry maturity. To further mitigate lighting-induced color distortion, RetinexNet was integrated to decompose, adjust, and reconstruct images by normalizing illumination and reflectance. Experiments demonstrated that SRR-Net achieved 0.918 mAP@50 for segmentation and operated at 210.3 FPS based on YOLOv11, while SRR-Net with RetinexNet attained 0.898 mAP@50 and 43.39 FPS. Though slightly lower in precision than other methods, both significantly improved ripeness accuracy, with mean absolute errors (MAE) of 0.040 and 0.037, representing 68.75 % and 71.09 % improvements over conventional Mask R-CNN approaches. Field experiments further demonstrated that SRR-Net and SRR-Net with RetinexNet achieved superior performance in ripeness regression. However, their detection and segmentation performance showed limited adaptability to real orchard conditions due to the characteristics of the LightStrawberry dataset. Overall, both models outperformed the standard YOLOv8/v11 baselines but were slightly inferior to Mask R-CNN. This work provides a robust solution for strawberry-harvesting robotics, enabling reliable ripeness assessment in challenging field environments.

Why it matches plant phenotyping methods画像からイチゴの成熟度を定量推定する回帰手法、照明補正、データセットを開発・評価しており、植物形質取得が中心である。

abstractpresents a novel light-resilient vision-based ripeness regression method

Automatic pixel-level annotation for plant disease severity estimation →

Plant disease adversely impacts food production and quality. Alongside detecting the disease, estimating its severity is important in managing the disease. Artificial intelligence deep learning-based techniques for plant disease detection are emerging. Unlike most of these techniques, which focus on disease recognition, this study addresses various plant disease-related tasks, including annotation, severity classification, lesion detection, and leaf segmentation. We propose a novel approach that learns the disease symptoms, which are then used to segment disease lesions for severity estimation. To demonstrate the work, a dataset of barley images was used. We captured the images of barley plants inoculated with diseases on test-bed paddocks at various growth stages. The dataset was automatically annotated at a pixel level using a trained vision transformer to obtain the ground truth labels. The annotated dataset was applied to train salient object detection (SOD) methods. Two top-performing lightweight SOD models were used to segment the disease lesion areas. To evaluate the performance of the SODs, we have tested them on our dataset and several other datasets, including the Coffee dataset, which has expert pixel-level labels that were unseen during the training step. Several morphological and spectral disease symptoms, including those akin to the widely used ABCD rule for human skin-cancer detection, i.e., asymmetry (A), border irregularity (B), colour variance (C), and diameter (D), are learned. To the best of our knowledge, this is the first study to incorporate these ABCD features in plant disease detection. We further extract visual and texture features using the grey level co-occurrence matrix (GLCM) and fuse them with the ABCD features. For the coffee dataset, our method achieved 82+% detection accuracy on the severity classification task. The results demonstrate the performance of the proposed method in detecting plant diseases and estimating their severity.

Why it matches plant phenotyping methods植物病害の病変を画像から抽出し、病害重症度という植物状態を推定する手法の開発・評価が中心であるため。

titleAutomatic pixel-level annotation for plant disease severity estimation

Deep learning framework for timely detection and classification of chili leaf diseases and pests. →

Food security continues to be a significant challenge the world over, with crop production becoming increasingly threatened by crop diseases and pest infestations. In the case of chili production, farmers often suffer significant yield loss and economic insecurity due to the unpredictable nature of both of these problems. Current pest control options (agrochemical and organic methods alike) have not reliably been enough for timely and effective control and demonstrate the importance of early and effective pest and disease identification processes. To solve this problem, the present work describes a deep learning implementation that leverages advanced YOLO-based architectures for the detection and classification of leaf diseases and pest on chili plant leaves. The present work trains and evaluates three models (YOLOv5, YOLOv7, and YOLOv8) on a newly created and balanced dataset of over 28,800 images combining 20 total classes of pest and leaf diseases. The dataset was supplemented by preprocessing the images and conducting an augmentation process to create a total of 32,000 images for training to generate reliable models. The results from the experiments found that YOLOv8 provided the best baseline performance of 95.1% mean Average Precision (mAP), while YOLOv5 had an mAP of 86.1%, and YOLOv7 had an mAP of 67.5%. An additional enhancement in the construction of a Modified YOLOv8 hybrid model—reflecting all advantages of YOLOv5, YOLOv7, and YOLOv8—achieved a highest mAP of 99.5% to be the most effective model in this study, the results suggest that the newly proposed Modified YOLOv8 framework, is highly accurate and reliable for the early detection of pests and diseases in chili, and is helpful to improve sustainable agricultural practices, mitigate crop losses, and increase global food security.

Why it matches plant phenotyping methodsチリ葉の病害を画像から検出・分類する深層学習手法の開発、比較評価、データセット構築が研究の中心であり、植物の病害状態を直接推定するため対象範囲に該当する。

abstractthe present work describes a deep learning implementation that leverages advanced YOLO-based architectures for the detection and classification of leaf diseases and pest on chili plant leaves.

Morphological diversity of pollen and spores in a human-impacted highland forest-agriculture mosaic in northern Thailand. →

Pollen and spore morphology provides essential taxonomic reference data for floristic and environmental studies in tropical regions, where modern comparative datasets remain limited. This study documents the morphological characteristics of pollen and spores recovered from a shallow soil profile in a degraded mixed deciduous forest within Sri Nan National Park, northern Thailand. Using a non-acetolysis extraction protocol and systematic sub-sampling of a 30-cm profile, pollen and spores representing 37 plant families were identified, including lycophytes, bryophytes, monilophytes, gymnosperms, and angiosperms. Spore-producing taxa, particularly monilophytes, dominate the assemblage, while angiosperm pollen includes both arboreal and non-arboreal elements. More than 100 morphotypes are described based on aperture type, exine ornamentation, size, and symmetry, supported by high-resolution photomicrographs and standardized morphotype descriptions. The resulting dataset expands the regional palynological reference framework for northern Thailand and tropical Southeast Asia and supports consistent taxonomic identification in palynological, floristic, and comparative paleoecological studies, particularly in human-impacted forest-agriculture mosaics.

Why it matches plant phenotyping methods植物の花粉・胞子形態を標準化して記載し、高解像度画像を含む再利用可能な地域参照データセットを構築しており、形態取得・記述が研究の中心です。

abstractMore than 100 morphotypes are described based on aperture type, exine ornamentation, size, and symmetry, supported by high-resolution photomicrographs and standardized morphotype descriptions.

PointNeXt-DBSCAN: a hybrid point cloud deep learning framework for multi-stage cotton leaf instance segmentation. →

This study addresses the challenge of organ-level instance segmentation in cotton point clouds, which arises from significant morphological variations and leaf occlusion across growth stages. To achieve high-precision leaf extraction, a hybrid framework integrating PointNeXt and DBSCAN is proposed. A dataset containing 1,065 cotton plants from seedling to boll-opening stages was constructed via multi-view image reconstruction and augmented through random rotation and scaling. Methodologically, a two-stage pipeline was designed: semantic segmentation was first performed using the PointNeXt network, where its residual MLP blocks enhanced edge and local feature learning; instance segmentation was then conducted by applying density-adaptive DBSCAN clustering to the semantic results, effectively mitigating over-segmentation in emerging leaves. Experimental results indicate that the semantic segmentation achieved an mIoU of 0.9846, representing a 7.2% improvement over PointNet++. The subsequent instance segmentation attained an ARI of 0.983, reduced the over-segmentation rate by 63%, and maintained an error below 3% for leaves smaller than 5 cm 2 . The framework provides reliable technical support for the automated extraction of key phenotypic traits such as leaf area index and leaf inclination distribution.

Why it matches plant phenotyping methods綿花葉の点群から器官レベルの葉を自動抽出するセグメンテーション手法を開発・評価し、葉面積指数や葉傾斜分布などの表現型形質への応用を示しているため、方法が中心的である。

abstractTo achieve high-precision leaf extraction, a hybrid framework integrating PointNeXt and DBSCAN is proposed.

Flow cytometry protocols, relative genome size and ploidy levels for 1104 species of non-apomictic angiosperms from the Eastern Alps - a community resource based on the screening of 45,000 samples →

Flow cytometry provides a reliable and fast method for estimating genome size and ploidy levels in plants. Until recently, most studies employed fresh tissues, which limits the use of the method with samples from remote areas or when an extremely high number of samples needs to be processed in a short time. Although there is growing evidence that silica-dried material can be used for ploidy estimation in some taxa, no flora-wide study has been available so far. Here, we provide methodological aspects of an unprecedented study exploring ploidy variation of non-apomictic angiosperms in the Eastern Alps. We have analysed ca. 45,000 silica-dried samples of 1135 species using flow cytometry with DAPI as stain. We were able to obtain ploidy level information from 1104 (97%) of species. The unsuccessful species included succulent plants of the family Crassulaceae (genera Jovibarba, Rhodiola, Sedum, Sempervivum), the achlorophyllous parasitic or mycoheterotrophic genera Orobanche and Hypopitis, and a handful of others. About 80% of samples were successfully analysed using a single universal protocol and leaf tissue, while in the remaining species the use of alternative tissues (such as petioles or flowers) and/or protocol modifications were needed (targeting composition of buffers, duration of fixation or staining time or use of alternative buffers). A total of 377 species (34%) included polyploid cytotypes and 179 (16%) species were ploidy-variable. As a community resource, we provide relative genome sizes and ploidy assignments of 1332 cytotypes retrieved from 1104 species along with methodological details (e.g. buffers, standards, analysed plant organs, histogram quality). We believe that this dataset will facilitate future research in particular species as well as in flora-wide investigations of ploidy level variation of the Central European flora in general. We are confident that novel cytotypes of many species will be discovered in other geographic areas, and we would be delighted if the present dataset could serve the botanical community for comparison.

Why it matches plant phenotyping methods植物のゲノムサイズ・倍数性を推定するフローサイトメトリー法の大規模な適用と、組織・バッファー・固定・染色条件の改良を中心に扱い、再利用可能なデータセットも提供しているため。

abstractFlow cytometry provides a reliable and fast method for estimating genome size and ploidy levels in plants.

Advancing Image Segmentation Techniques for Strawberry Detection in Vision-Based Agricultural Robotics →

Image segmentation is a fundamental component of vision-based agricultural robotics, enabling accurate fruit localization, disease detection, and automated harvesting. However, real-world strawberry fields present significant challenges due to irregular fruit morphology, dense foliage occlusions, variable ripeness, and strong illumination variability. Moreover, segmentation models trained on a single dataset often fail to generalize across domains, limiting their practical deployment. This paper presents a comprehensive benchmark of classical computer vision methods, convolutional neural networks, instance-based models, and transformer-based architectures across three heterogeneous public strawberry datasets: Db1 (instance segmentation), Db2 (lesion segmentation), and Db3 (semantic segmentation). A unified preprocessing and evaluation framework is adopted to ensure fair comparison using standard metrics, including Intersection-over-Union (IoU), Dice coefficient, Precision, and Recall. Extensive in-domain experiments demonstrate that deep learning models significantly outperform classical approaches, with U-Net and SegFormer achieving IoU values above 0.95 on Db1 and up to 0.83 on Db3. Cross-domain zero-shot evaluations reveal a substantial generalization gap, with U-Net suffering IoU drops of up to 100\%, while SegFormer consistently exhibits improved robustness and reduced cross-domain degradation across most transfer scenarios. To our knowledge, these results establish the first systematic multi-dataset benchmark for strawberry segmentation under domain shift, highlighting the importance of transformer-based architectures for robust agricultural perception and providing practical insights for real-world robotic deployment.

Why it matches plant phenotyping methodsイチゴの病斑・果実を画像から分割する手法を複数データセットで比較・ベンチマークし、ドメインシフト下の性能を評価しているため、植物の病害状態・器官形態の取得が中心である。

abstractThis paper presents a comprehensive benchmark of classical computer vision methods, convolutional neural networks, instance-based models, and transformer-based architectures across three heterogeneous public strawberry datasets

A Multi-Scale Perception-Enhanced Lightweight Network with Knowledge Distillation for Rice Leaf Disease Detection. →

Rice is a critical crop for global food security and economic stability. However, various diseases, including rice blast and bacterial leaf blight, pose significant threats to rice cultivation. Existing methods for detecting rice leaf diseases suffer from low efficiency and limited generalization capability. These methods are incapable of capturing variations of disease characteristics across different growth cycles. Therefore, a lightweight detection model named lightweight knowledge distillation YOLO (LWKD-YOLO) is proposed. The convolutional layers in the YOLOv8 network are replaced with the ADown module. This change significantly reduces computational load while improving detection accuracy. A lightweight detection head, termed the lightweight shared re-parameterizable convolutional detection head (LSRP-Head), was designed. It incorporates group normalization RepConv, further reducing computational complexity while enhancing multi-scale perception capabilities. Furthermore, based on the improved ADown module and LSRP-Head, the YOLOv8x model is employed as a teacher model for inter-channel correlation knowledge distillation. This effectively enhances the ability to learn complex rice leaf disease features. The effectiveness of the proposed method was verified through ablation and comparative experiments on the constructed rice leaf disease dataset. Compared with the baseline model, LWKD-YOLO increases mAP@50 by 1.4%, reduces the number of parameters by 1.3M, and lowers FLOPs by 3.1G. As a result, the proposed model enables efficient rice leaf disease detection in complex environments, demonstrating notable economic and practical significance.

Why it matches plant phenotyping methodsイネ葉の病害状態を画像から検出する軽量な深層学習手法を開発し、データセット上で比較・アブレーション検証しており、植物表現型取得が中心である。

abstracta lightweight detection model named lightweight knowledge distillation YOLO (LWKD-YOLO) is proposed.

High-Performance Deep Learning Techniques for Plant Disease Detection: Performance Analysis, Validation, and Applications →

The early detection of plant diseases is an indispensable task to improve crop yields and production quality. Crop disease observations by experienced pathologists are difficult and might take a long time. Therefore, deep learning (DL) techniques have been utilized to present an automated detection technique that could accurately and timely detect plant diseases. Several DL models in the literature were proposed, but no paper conducted a comparative study between those models to determine which of them was the best alternative for this task. Therefore, twenty-one DL models are compared in this review paper to show which of them could achieve better classification accuracy when applied to detect plant diseases. Three publicly available datasets, namely PlantVillage, Tomato Leaves, and Groundnut Plant Leaf, are used to assess the performance of those models under five different performance metrics, such as accuracy, precision, recall, F1-score, and area under curve (AUC). The extensive experiments conducted in the same environments under the same number of epochs and batch size for all models show that EfficientNetB0 is the best for both PlantVillage and Tomato Leaves datasets, with a classification accuracy of around 99% and 98%, respectively, and ResNet152 is the best for the Groundnut Plant Leaf dataset, with a classification accuracy of 99.7%.

Why it matches plant phenotyping methods植物葉の病徴を画像から分類する深層学習手法を21モデルで比較・評価しており、植物病害状態の表現型推定と技術ベンチマークが中心である。

abstractTherefore, twenty-one DL models are compared in this review paper to show which of them could achieve better classification accuracy when applied to detect plant diseases.

Intelligent Evaluation of Rice Resistance to White-Backed Planthopper (Sogatella furcifera) Based on 3D Point Clouds and Deep Learning →

Accurate assessment of rice resistance to Sogatella furcifera (Horváth) is essential for breeding insect-resistant cultivars. Traditional assessment methods rely on manual scoring of damage severity, which is subjective and inefficient. To overcome these limitations, this study proposes an automated resistance evaluation approach based on multi-view 3D reconstruction and deep learning–based point cloud segmentation. Multi-view videos of rice materials with different resistance levels were collected over time and processed using Structure from Motion (SfM) and Multi-View Stereo (MVS) to reconstruct high-quality 3D point clouds. A well-annotated “3D Rice WBPH Damage” dataset comprising 174 samples (15 rice materials, three replicates each, 45 pots) was established, where each sample corresponds to a reconstructed 3D point cloud from a video sequence. A comparative study of various point cloud semantic segmentation models, including PointNet, PointNet++, ShellNet, and PointCNN, revealed that the PointNet++ (MSG) model, which employs a Multi-Scale Grouping strategy, demonstrated the best performance in segmenting complex damage symptoms. To further accurately quantify the severity of damage, an adaptive point cloud dimensionality reduction method was proposed, which effectively mitigates the interference of leaf shrinkage on damage assessment. Experimental results demonstrated a strong correlation (R2 = 0.95) between automated and manual evaluations, achieving accuracies of 86.67% and 93.33% at the sample and material levels, respectively. This work provides an objective, efficient, and scalable solution for evaluating rice resistance to S. furcifera, offering promising applications in crop resistance breeding.

Why it matches plant phenotyping methods3D画像再構成と深層学習によってイネの害虫被害症状・被害重症度を定量化する手法を開発・検証しており、植物表現型取得が研究の中心である。

abstractthis study proposes an automated resistance evaluation approach based on multi-view 3D reconstruction and deep learning–based point cloud segmentation.

An Efficient Additive Kolmogorov-Arnold Transformer for Point-Level Maize Localization in Unmanned Aerial Vehicle Imagery →

High-resolution UAV photogrammetry has become a key technology for precision agriculture, enabling centimeter-level crop monitoring and point-level plant localization. However, point-level maize localization in UAV imagery remains challenging due to (1) extremely small object-to-pixel ratios, typically less than 0.1%, (2) prohibitive computational costs of quadratic attention on ultra-high-resolution images larger than 3000 x 4000 pixels, and (3) agricultural scene-specific complexities such as sparse object distribution and environmental variability that are poorly handled by general-purpose vision models. To address these challenges, we propose the Additive Kolmogorov-Arnold Transformer (AKT), which replaces conventional multilayer perceptrons with Pade Kolmogorov-Arnold Network (PKAN) modules to enhance functional expressivity for small-object feature extraction, and introduces PKAN Additive Attention (PAA) to model multiscale spatial dependencies with reduced computational complexity. In addition, we present the Point-based Maize Localization (PML) dataset, consisting of 1,928 high-resolution UAV images with approximately 501,000 point annotations collected under real field conditions. Extensive experiments show that AKT achieves an average F1-score of 62.8%, outperforming state-of-the-art methods by 4.2%, while reducing FLOPs by 12.6% and improving inference throughput by 20.7%. For downstream tasks, AKT attains a mean absolute error of 7.1 in stand counting and a root mean square error of 1.95-1.97 cm in interplant spacing estimation. These results demonstrate that integrating Kolmogorov-Arnold representation theory with efficient attention mechanisms offers an effective framework for high-resolution agricultural remote sensing.

Why it matches plant phenotyping methodsUAV画像から個体位置を抽出する手法を開発し、個体数と株間距離という植物群落形質を推定しており、データセット構築と技術評価も中心的である。

abstractTo address these challenges, we propose the Additive Kolmogorov-Arnold Transformer (AKT)

A Uav-based multisensor framework for legal industrial Cannabis monitoring and open-access dataset development. →