Plant Phenomics

Guava cultivation is considerably influenced by foliar and fruit diseases whose overlapping symptoms and environmental variability make accurate field-level diagnosis challenging. Numerous studies have been conducted to find efficient methods of diagnosing plant diseases, but most focus on image-level classification and do not include lesion localization or pixel-level segmentation of the images within a single framework of analysis. This study proposes a comprehensive framework for utilizing automated image analysis to classify guava leaf and fruit diseases at the image level, locate lesions, and segment lesions at the pixel level from multiple images of the same type of disease collected from various growing conditions. The dataset was enriched through three augmentation strategies including standard preprocessing, structured augmentation, and GAN-based synthetic image generation, expanding the effective training data to approximately 7,000 images, while a 5-fold cross-validation strategy guided model selection and final performance was assessed on a held-out test set. The experimental evaluation of multiple state-of-the-art Convolutional Neural Networks (CNNs) for the classification of guava leaf and fruit diseases indicated that the model generated using the ResNet50+DenseNet121 model fusion achieved the highest classification accuracy of 98.20%. For lesion detection and segmentation, YOLOv8-seg outperformed Mask R-CNN, achieving mAP@0.5 of 0.907 and 0.889, and mAP@0.5:0.95 of 0.783 and 0.769 for detection and segmentation, respectively, with a balanced precision-recall profile. The techniques of Explainable AI (XAI) were used to increase the transparency of this model by identifying areas in the image that are significant to the actual lesion. The framework was further designed with practical web-based deployment in mind, evaluating both lightweight and high-capacity models to balance computational efficiency against predictive accuracy. From this research, it was concluded that using model fusion, data augmentation, and segmentation-aware lesion detection would provide a solution for managing guava diseases effectively.

Image-based crop and pest recognition is considered useful for reducing the delay and cost of manual field scouting, therefore supporting timely intervention in precision-agriculture workflows. However, the real field imagery remains challenging due to the cluttered backgrounds, occlusions, illumination changes, and strong scale variation that are frequently observed across crops. The symptoms are often small or low-contrast, and pests may be partially hidden, which reduces the reliability when the setting is outside controlled environments. A unified multi-class crop-pest/condition recognition framework is presented, where a ResNet-50 backbone is utilized and enhanced with a Multi-Scale Contextual Attention (MSCA) module. The novelty is mainly considered to be achieved through the integration of explicit multi-scale contextual aggregation with lightweight joint channel and spatial attention by means of residual fusion, while the empirical evaluation was kept controlled under a fixed and reproducible protocol. A curated dataset of 21,404 field-style images covering 15 crop and pest/condition classes was compiled, and a leakage-aware fixed split with a held-out test set was adopted to support reproducibility. Augmentation was applied only to the training subset to improve robustness, although the validation data was not augmented in the same manner. On the held-out test set, balanced performance was achieved by the proposed approach, with about 0.93 accuracy and a macro-F1 score close to 0.94 being obtained, while established baselines such as EfficientNet, Vision Transformer, and attention-based CNN models were outperformed under identical evaluation settings. Controlled ablations were used to isolate the contribution of MSCA and augmentation under the same training configuration. These results indicate that lightweight multi-scale contextual attention is effective for crop and pest recognition under realistic field conditions, although some visually similar classes remained difficult.

Plant disease detection using deep learning is essential for precision agriculture, enabling early and automated crop health monitoring. This study proposes an end-to-end transfer learning pipeline, LeafyVGG-16, for multi-class classification of plant diseases and nutrient deficiencies using a tomato leaf dataset. The framework integrates data preprocessing, augmentation, and a VGG-16 backbone with a two-stage fine-tuning strategy. The proposed model is evaluated against CNN, DenseNet-121, Inception-V3, EfficientNetB0, and ResNet-50, achieving an accuracy of 0.93 with precision, recall, and F1-scores of 0.93, 0.90, and 0.92, respectively. These results demonstrate the effectiveness of transfer learning for fine-grained plant disease recognition. We further evaluate model robustness under adversarial cyber attacks to assess deployment reliability in agricultural systems. Under Fast Gradient Sign Method (FGSM) attacks ({epsilon} = 0.01- 0.05), the model shows an accuracy drop of 1%-7.5%, while Projected Gradient Descent (PGD) attacks ({epsilon} = 0.05, step size = 0.005, 10 iterations) produce similar degradation, highlighting the models vulnerability to adversarial perturbations. These findings highlight potential security and reliability risks in AI-based agricultural decision-making systems. Future work will focus on improving robustness and cyber-resilience and extending this framework to other crops for secure and context-aware deployment in resource-constrained environments.

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 (R2 = 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 SUMMARYSouthern 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.

Early, precise, and non-destructive stress detection is essential for maintaining crop productivity, particularly in high-density plant growth systems like controlled environment agriculture (CEA), where manual monitoring is often impractical. Using plant motion as a proxy for growth and plant health, we demonstrate a method for early, non-invasive stress detection through quantitative leaf-movement analysis in lettuce and five other CEA relevant crops. Leaf-movement dynamics under stress were imaged with a low-cost, scalable Raspberry Pi imaging setup and quantified using a repurposed open-source motion estimation algorithm; Tracking Rhythms in Plants (TRiP). Our system detected stress-induced changes in leaf-movement within 1 hour of stress, with the timing dependent on the nature of the stress. Sustained reductions in leaf-movement coincide with decreased biomass accumulation. This approach offers a non-invasive, rapid, scalable, and cost-effective solution for continuous crop monitoring, with potential for application in both terrestrial and space farming CEA systems. O_FIG O_LINKSMALLFIG WIDTH=200 HEIGHT=138 SRC="FIGDIR/small/732190v1_ufig1.gif" ALT="Figure 1"> View larger version (54K): org.highwire.dtl.DTLVardef@19ee20eorg.highwire.dtl.DTLVardef@b0804org.highwire.dtl.DTLVardef@3b3fa8org.highwire.dtl.DTLVardef@1d04026_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOGraphical abstract:C_FLOATNO Quantification of leaf-movement dynamics as a high-throughput proxy for plant physiological status, enabling early stress detection and timely intervention to mitigate yield penalties in CEA settings (image made with biorender.org). C_FIG

While whole genome sequencing captures millions of single nucleotide polymorphisms (SNPs) and hyperspectral imaging (HSI) enables non destructive plant phenotyping, integrating these modalities to link genotype to phenotype remains challenging due to their high dimensionality and non linearity. This study presents DeepPheno a deep learning framework that predicts SNP genotypes from HSI data, using model predictability as a proxy for genotype phenotype association. HSI data were acquired from 194 lettuce genotypes under field conditions. HSI data patches (20 x 20 pixels x 224 spectral bands) were used to train a hybrid CNN to predict the variant of a specific SNP. The framework was validated on SNPs with known phenotypic effects (anthocyanin, leaf serration, pale pigmentation), achieving high predictive performance (AUC ranging from 0.806 to 0.935), whereas models trained on randomly shuffled labels performed at chance (mean AUC {approx} 0.51). Extending the workflow to 50 randomly selected putatively neutral SNPs, most yielded low predictability, but two showed high performance (AUC > 0.76), suggesting uncharacterized genotype phenotype links. Explainable AI, including SHAP and Grad CAM, identified relevant spectral and spatial features driving these predictions, particularly the green and red edge wavelengths associated with pigment dynamics and leaf structure. These results establish a framework for understanding complex genotype phenotype interactions in plants and extracting these links from HSI data without predefining the exact trait values. It provides an avenue for high throughput trait discovery and description and extends the integration of image based phenomics with plant genetics.

High-throughput 3D multispectral plant phenotyping platforms generate large volumes of point cloud files, but trait extraction is typically performed by sensor-bundled software whose internal algorithms are not publicly documented, which limits reproducibility and integration into custom research pipelines. Here we present PhytoScan3D, an open-source Python pipeline that extracts morphological and spectral phenotypic traits, spanning plant height, 3D leaf area, digital biomass, convex hull volume, leaf inclination, canopy geometry, NDVI, hue, and vegetation indices, from both PLY and PCD point cloud files generated by Phenospex PlantEye F500 and F600 sensors, and is portable to point clouds from any acquisition platform. PhytoScan3D was validated against HortControl (PhenoSpex) ground-truth measurements on 936 barley (Hordeum vulgare) pot-date observations from the growth chamber trial (20 Norwegian cultivars, 12 scan dates, Septemenr 2025 to January 2026), achieving Pearson r = 0.913 to 0.999 and ratio approximately 1.000 for Plant Height Max, 3D Leaf Area, and NDVI Average. A vectorised mesh face filtering implementation achieved a 120x speed improvement, increasing valid 3D Leaf Area coverage from 0.6% to 100% of files. Cross-format validation on 223 PlantEye F600 PCD files from the ICRISAT LeasyScan platform (four legume species: mungbean, cowpea, lima bean, and common bean; 1,523 plant observations) yielded r = 0.884 against independent cuboid annotation heights. The systematic positive bias (mean +27.2 mm, ratio = 1.44) is attributable to PhytoScan3D computing height from raw point cloud Z-range while cuboid annotations are fitted to segmented plant points only, with the offset consistent across all four species (per-species r = 0.880 to 0.888). Cross-dataset processing of 1,180 PLY files from the Crops3D benchmark (8 species, 3 acquisition methods) confirmed zero extraction errors. PhytoScan3D is available at "github.com/kovimallik/phytoscan3d" under the MIT licence and processes 1,651 files across three independent datasets in under 12 minutes on GPU hardware. HighlightsO_LIPhytoScan3D is the first open-source Python pipeline for batch extraction of phenotypic traits, including plant height, 3D leaf area, digital biomass, convex hull volume, leaf inclination, NDVI, and excess green index, from both PLY and PCD point cloud files generated by Phenospex PlantEye sensors. C_LIO_LIPrimary validation against HortControl ground-truth measurements on 936 barley pot-date observations achieved Pearson r = 0.913-0.999 for Plant Height Max, 3D Leaf Area, and NDVI Average. C_LIO_LIA 120x computational speedup in mesh face filtering (vectorised NumPy vs. set-based loop) increased the coverage of valid 3D Leaf Area extraction from 0.6% to 100% of files. C_LIO_LICross-format validation on 223 PlantEye F600 PCD files from ICRISAT LeasyScan (four legume species, 1,523 plants) achieved r = 0.884 against independent cuboid annotation heights. The systematic +27.2 mm bias reflects a methodological difference (raw Z-range vs. soil-segmented annotations), is consistent and predictable across all four species (per-species r = 0.880-0.888), and is correctable by a single linear factor. C_LIO_LICross-dataset processing of 1,180 PLY files from the Crops3D benchmark (8 species, 3 acquisition methods) confirmed zero extraction errors. C_LIO_LISignificant scan-unit variation was detected for Plant Height Max (F = 5.71, p < 0.001, 2 = 0.138) and Canopy Width X (F = 6.32, p < 0.001, 2 = 0.150), demonstrating the biological utility of extracted traits. C_LI

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.

Quantifying root traits such as root length (RL) and root surface area (RSA) from minirhizotron imagery is a valuable approach for overcoming the phenotyping bottleneck that limits understanding and improvement of crop productivity, resource use efficiency and resilience in field experiments. However, current approaches remain labor-intensive, and deep learning (DL) methods suffer from limited generalization ability. We present RootQuant, an end-to-end DL model that simultaneously predicts RL and RSA directly from minirhizotron images using only whole-image trait values as supervision, thereby eliminating the need for pixel-level annotations. The models generalization ability was evaluated across species and fine-tuning configurations. The practical applicability of the model was further assessed under field conditions by converting image-derived RL estimates into volumetric root length density (vRLD). Using 118,191 maize and soybean images collected between 2009 and 2020, RootQuant trained on both species achieved an R2 of 0.90 and an RMSE of 2.9 mm for RL, and an R2 of 0.88 and an RMSE of 4.2 mm2 for RSA. The same mixed-species model generalized strongly across species, yielding an 8% relative improvement in R2 and a 30% lower RMSE on maize compared with the same architecture trained on a single species and applied zero-shot. Image-derived RL predictions converted to vRLD showed the expected depth-dependent decline in vRLD, as was also found by coincident destructive quantification of roots washed out of soil cores. By providing a generalist backbone model trained on a large dataset from two major crop species, RootQuant enables high-throughput simultaneous estimation of two relevant root traits directly from raw imagery without task-specific fine-tuning, thereby accelerating in situ root system analysis and phenotyping applications.

Accurate segmentation of plant disease symptoms is essential for crop monitoring and phenotyping, yet it typically requires costly pixel-level annotations. Weakly supervised semantic segmentation (WSSS) alleviates this burden using image-level labels, but its performance depends on the quality of spatial priors such as class activation maps (CAMs). We investigate whether text-guided segmentation with the Segment Anything Model 3 (SAM3) can serve as an alternative weak supervision signal. Three pseudo-mask generation strategies are compared: (i) CAMs refined with SAM or SAM3, (ii) zero-shot text-guided SAM3, and (iii) a hybrid approach combining weak spatial cues with text prompts. The resulting pseudo-masks are used to train a DeepLabV3 model. Text guidance alone matches or outperforms conventional WSSS, achieving up to 0.46 IoU without spatial supervision and 0.61 IoU on a public dataset, although performance is sensitive to text prompt formulation. The hybrid strategy improves robustness, reaching 0.50 IoU on the primary dataset and 0.58 IoU on the additional dataset while reducing prompt sensitivity. Overall, text guidance is a promising alternative to conventional weak supervision, while hybrid approaches provide a more robust solution for plant disease segmentation.

Accurate and reproducible assessment of foliar disease severity is essential for evaluating the performance of heterogeneous plant communities and understanding host-pathogen interactions. However, traditional visual scoring methods remain subjective, with limited precision, and difficult to scale in large phenotyping experiments. Here, we present a semi-automated image analysis workflow designed to quantify multiple foliar disease symptoms simultaneously on wheat flag leaves sampled from varietal mixtures. The workflow combines three methodological components: (i) a standardized protocol for leaf sampling and imaging, (ii) supervised machine learning segmentation using Random Forest implemented in Ilastik to classify multiple symptoms (powdery mildew and yellow rust), and (iii) a graphical user interface facilitating pipeline deployment by non-specialist operators. To evaluate the influence of image representation on classification performance, four color spaces (RGB, HSV, HLS, LAB) were systematically compared. The approach was validated using images of durum wheat flag leaves collected from a field experiment assessing eight-way varietal mixtures under natural fungal pressure. Cross-validation against manually annotated images demonstrated high segmentation accuracy across all symptom. Comparison among color spaces revealed only minor differences in performance. Overall, this workflow offers a cost-effective, annotation-efficient and reproducible alternative to deep learning approaches, leveraging open-source and actively maintained tools while requiring limited training data and enabling objective, reproducible and scalable disease phenotyping.

Accurate quantification of leaf lesion severity is essential for plant disease research and phenotyping but is often limited by subjective visual scoring and time-intensive manual image analysis. We present LIME, a fully automated, open-source image analysis pipeline for high-throughput quantification of leaf lesions from disease assay images. LIME integrates zero-shot leaf segmentation using the Segment Anything Model with a convolutional neural network for lesion area estimation. Applied to Arabidopsis thaliana leaves infected with Sclerotinia sclerotiorum, the proposed approach achieved a mean absolute percentage error of 12.9%, comparable to observed intrarater variability in manual scoring. Stratified evaluation across lesion-size groups demonstrated consistent prediction accuracy for small, intermediate, and large lesions, and comparative analysis showed that the deep learning-based model substantially outperformed color-based baseline methods. Under GPU-accelerated execution, LIME processed complete assays containing approximately 200 leaves in 15 minutes, representing an approximate 13-fold reduction in processing time relative to manual annotation. Together, these results indicate that LIME enables objective, reproducible, and scalable quantification of leaf lesion severity in standardized plant pathology assays. The pipeline is released as an open-source tool to support quantitative phenotyping studies.

Potato crop is highly vulnerable to abiotic stresses like salinity and low nutrient availability. Rapid identification of stress-resilient genotypes is therefore essential for breeding, yet conventional phenotyping is often slow, space-demanding and expensive. We present LOCOPOTS -- a LOw-COst high-throughput screening platform for in vitro POTatoes under abiotic Stress -- which combines individual in vitro plant culture, low-cost RGB imaging and machine-learning-based automatic segmentation using a trained model of a convolutional neural network, based on U-Net architecture. LOCOPOTS enabled the automated extraction of growth, colour, and vegetation-index traits and demonstrated robust performance across independent phenotyping rounds. We screened 30 potato varieties under control, low-nutrient and saltinity conditions, identifying contrasting growth and physiological responses. Integrated traits such as final area and height, Area_AUC and height_AUC, together with GLI, Chol, cive and chlorophyll fluorescence parameters, discriminated genotype performance under stress. Metabolic profiling further revealed genotype-specific reprogramming in carbon and nitrogen metabolism under low nutrition and salt stress, including changes in fructose, myo-inositol, {beta}-aminobutyric acid, {gamma}-aminobutyric acid, proline, and certain polyamines, identifying them as specific chemical biomarkers of plant stress responses. LOCOPOTS provides a scalable, affordable and space-efficient platform for early screening of potato genetic diversity and identification of candidate traits associated with stress resilience.

Kernel composition traits in maize, including protein accumulation, are of broad interest. The amount of the most abundant proteins in maize endosperm, the -zeins, can vary dramatically among genotypes and in response to soil nitrogen supply. Targeted reductions in -zein accumulation can improve nitrogen utilization and the nutritional quality of maize grain but have traditionally required expensive and destructive phenotyping methods. The Floury2-RFP (Fl2-RFP) reporter gene enables rapid, non-destructive visualization of -zein accumulation in individual maize kernels under white light. This feature is due to the high expression level programmed by the Fl2 promoter, the stability of zein proteins, and the use of monomeric RFP, which emits fluorescence without the need for multimerization. This study aimed to develop a method to quickly document and quantify Fl2-RFP accumulation using camera or smartphone images of either ears or shelled kernels. Results show images of shelled kernels processed with FIJI software capture the Fl2-RFP reporter phenotype better than images of ears. Fl2-RFP confirms the strong maternal control of -zein accumulation and, like grain protein concentration, responds to soil nitrogen supply. The Fl2-RFP phenotyping pipeline effectively quantified Fl2-RFP accumulation by color features from both camera and smartphone images. Smartphone imaging of Fl2-RFP in a diverse population of inbreds followed by elastic net regression of extracted image features predicted kernel protein concentration, as measured by near-infrared spectroscopy, with moderate accuracy (R2 = 0.68, MAE = 0.76, RMSE = 0.93). The spectral features that were most predictive of kernel protein concentration varied depending on whether the background endosperm color was white or yellow. The integrated analysis of Fl2-RFP intensity and grain protein concentration indicates genetic variation for kernel protein accumulation and N-responsiveness that is distinct from the well-studied -zeins. Our findings highlight the Fl2-RFP reporter gene as a valuable tool for investigating the genetic complexity of grain protein concentration and associated traits in maize.

The Digital Pedon (DP) is an open-source Python framework that represents a soil profile as a continuously updated digital twin, bridging three persistent gaps in soil science: disconnected models and observations, cross-database interoperability, and the inference gap between raw sensor signals and agronomically meaningful variables. Integrating real-time sensor streams, model-based solver chains (Model-Zoo), GLOSIS-compliant ontology mapping, and a novel LLM agentic interface layer enabling natural language soil queries, the DP supports applications spanning precision agriculture, digital soil mapping, and environmental sustainability assessment. Four proof-of-concept experiments confirm automatic profile initialisation fidelity, solver chain consistency, ontology compliance, and user-defined solver extensibility.

In wheat, the pre-heading stage determines spikelet formation, floret fertility, and canopy development, making it a critical window for early stress detection and yield potential. The genetic basis of pre-heading canopy development in wheat has remained constrained by the conventional phenotyping due to the low temporal resolution. Here, we quantified the rate of vegetation index gain (RVIs) during tillering to heading stages using UAV-mounted multispectral sensor in 196 spring wheat cultivars representing 112 years of breeding history. RVIs were calculated using six vegetation indices for consecutive two growing seasons, and genome wide association study (GWAS) was performed on RVIs, grain yield (GY) and thousand grain weight (TGW) using a wheat 37K SNP array. RVIs showed significant positive correlations with grain yield (r=0.28-0.43; p<0.001) and consistently increased in the modern cultivars compared to old cultivars. This indicated that resource remobilization during pre-heading canopy development significantly contributed to GY during modern wheat breeding. GWAS identified 67 loci, including 12 Group-I loci associated only with RVIs, and 18 Group-II loci associated with both RVIs and yield traits. Two stable loci on chr1B and chr5D consistently increased GY and RVIs across environments, and the tag SNPs were converted to selectable KASP markers. The allelic distribution on global wheat collection of [~]3000 accessions showcased that favorable alleles on both loci were dominant in cultivars compared to landraces. Similarly, favorable alleles showed more frequency in winter type than spring type. Across breeding eras both alleles showed increasing trend with chr5D reaching near fixation and chr1B remaining partially enriched in modern cultivars. Our work on capturing pre-heading canopy development, discovery of two stable loci underpinning yield and RVIs, and development of KASP markers provided a strong foundation to HTP assisted genetic dissection of GY and facilitated the understanding of canopy dynamics and yield formation.

Achieving high-throughput and precise phenotypic quantification and imaging modalities of stomatal and epidermal cells across diverse species remains a primary bottleneck in elucidating the mechanisms of stomatal dynamics, epidermal patterning, and environmental adaptation of plants. Here, we developed EpiReasoner, an artificial intelligence framework comprising a vision module, EpiVision, and a knowledge-based reasoning module, EpiBrain, for the quantitative phenotypic analysis and domain-specific knowledge reasoning of stomatal complexes and pavement cells in plants. Operating across bright-field, scanning electron microscopy, and differential interference contrast modalities, EpiVision achieves precise instance segmentation in various monocotyledonous, dicotyledonous, and fern species. Its performance significantly surpasses current state-of-the-art models. Moreover, we defined 23 quantitative indices describing stomatal cell morphology and spatial distribution. For domain-specific tasks such as phenotype prediction, genotype deduction, and molecular mechanism reasoning, EpiBrain demonstrates a human preference rate significantly higher than that of general-purpose large language models, including GPT-5 and Claude Sonnet 4. The application of EpiReasoner to phenotypic data of stomatal density derived from a tomato natural population of 170 accessions successfully identified a major quantitative trait locus on chromosome 8. The candidate gene, SKP1-interaction partner 19L (SKIP19L), encoding an F-box family protein, exhibited severe allele frequency drift during tomato domestication, which is highly consistent with the adaptive trend of reduced stomatal density under artificial selection. EpiReasoner provides a novel paradigm that unifies visual phenomics and knowledge-driven reasoning for the biology of stomata and pavement cells, thereby significantly accelerating scientific discovery in plant science.

PremiseSeed size and morphology are critical traits in agriculture, ecology, and genetics, but high-throughput quantification of these traits is often limited by labor-intensive manual measurements or expensive, platform-specific imaging software. Methods and ResultsWe developed SeedMeasure, a lightweight, open-source, and cross-platform command-line tool written in Python that automates the measurement of seed area, length, and width from images. Using a simple imaging setup, the program processes images by correcting for perspective skew, filtering debris, and exports quantitative data alongside quality-check images. We validated SeedMeasure across nine diverse species, ranging from small Arabidopsis thaliana seeds to large Zea mays kernels. The tool quickly handles images using multithreading and demonstrates high reproducibility, yielding low coefficients of variation across repeated runs. ConclusionsCompared to existing software, SeedMeasure is free, offers faster processing through parallel computing, and provides standalone executables that require no programming dependencies. SeedMeasure offers an accessible, cost-effective, and high-throughput approach for rapid phenotypic profiling, making advanced seed morphological analysis available to researchers without specialized laboratory hardware.

Waterlogging is a major constraint on barley productivity, yet its dynamic, multi-phase nature makes it challenging to dissect using traditional phenotyping approaches. High-throughput phenotyping (HTP) platforms address this by enabling temporal, multi-sensor imaging of large populations, but generate complex datasets that demand new analytical frameworks. Here, we imaged 230 barley accessions over 14 days of waterlogging stress and seven days of recovery using visible, chlorophyll fluorescence, and hyperspectral sensors. Explainable AI was applied to classify stress responses into early stress, late stress, and recovery phases, achieving 86% classification accuracy, and to identify the hyperspectral indices most informative for each phase. Water index (WATER1) and structure insensitive pigment index (SIPI) emerged as primary predictors of stress response. Longitudinal genome-wide association studies (GWAS), using a treatment-by-marker interaction model, identified 236 significant loci across 12 linkage disequilibrium blocks, implicating candidate genes involved in oxidative stress regulation, transcriptional control, and auxin transport. MYB transcription factors were consistently identified across all stress phases, underscoring their central role in waterlogging adaptation. To support interpretation of longitudinal GWAS results, we developed 3D-QTLVis, an interactive visualisation tool that extends Manhattan plots across time, enabling clearer identification of dynamic genomic regions underlying stress tolerance.

1Non-structural carbohydrates (NSCs) are central to plant carbon allocation and physiological regulation, yet their quantification typically relies on destructive biochemical assays that lack spatial resolution. Here, we developed a shortwave infrared (SWIR) hyperspectral imaging workflow for non-destructive estimation and spatial reconstruction of starch-associated variation in strawberry leaves. The workflow combined automated hyperspectral segmentation, spectral preprocessing, Partial Least Squares Regression (PLSR), and constrained wavelength selection. Sample-level spectra extracted from 114 strawberry leaf samples grown across three different metabolic conditions were paired with destructive starch measurements and used to train models across the 900-1750 nm spectral range. A constrained greedy band-selection strategy revealed that predictive performance approached a plateau at approximately 12 wavelengths, indicating substantial spectral redundancy within the full hyperspectral dataset. The final reduced-band model achieved a cross-validated coefficient of determination (R2) of 0.771 {+/-} 0.066 and a root mean squared error (RMSE) of 0.743 {+/-} 0.098 mg g-1 fresh weight using repeated stratified 5-fold cross-validation. Pixel-wise application of the final model generated spatial starch-associated maps that preserved pronounced intra-leaf heterogeneity, including vein-associated spatial structure. These results demonstrate that starch-associated spectral information can be reconstructed from a constrained reduced-band SWIR framework while retaining sufficient predictive performance for spatial mapping. The identified wavelength reduction supports the feasibility of deployable multispectral systems for non-destructive carbohydrate sensing in plant phenotyping applications.