Stomatal trait analysis is essential for optimizing crop photosynthesis and transpiration, yet deep learning studies have focused mainly on monocotyledons, leaving dicotyledonous crops such as melon (Cucumis melo L.) understudied. To bridge this gap, we established a dedicated melon stomatal dataset comprising 5,708 training images, 1,631 validation images, and 815 test images. On this basis, we developed an improved Mask R-CNN framework using Vision Transformer (ViT) as the backbone. Specifically, standard Multi-Head Attention (MHA) was replaced with Dynamically Composable Multi-Head Attention (DCMHA), which enhances information exchange across attention heads and alleviates the low-rank limitation of conventional attention. In addition, a modified effective Squeeze-and-Excitation (eSE) module was incorporated into the Feature Pyramid Network (FPN) to strengthen channel dependency modeling and multi-scale feature representation. On the melon dataset, the proposed model achieved a mean average precision (mAP) of 72.40 ± 0.09%, with AP50 and AP75 of 91.93 ± 0.14% and 84.59 ± 0.19%, respectively. Repeated-run statistical analyses showed that eSE significantly and consistently improved the main detection metrics across backbones, whereas DCMHA provided a more moderate gain within the ViT-based setting, with clearer support for AP50 than for mAP or AP75 under the baseline FPN setting. Overall, the combined configuration remained among the top-performing models for stomatal instance segmentation. Ellipse fitting further enabled automated quantification of stomatal length, width, count, area, and circumference, showing strong agreement with manual measurements (Pearson r = 0.978). The model also showed preliminary transferability to cucumber, watermelon, pumpkin, and loofah, with an average species-specific R² of 0.86, although each species was evaluated on a limited sample set.
Why it matches plant phenotyping methodsメロンの気孔形質を画像から自動抽出するデータセット、改良Mask R-CNN、インスタンスセグメンテーション、楕円フィッティングを開発・検証しており、植物フェノタイピング手法が研究の中心である。
abstractwe established a dedicated melon stomatal dataset comprising 5,708 training images, 1,631 validation images, and 815 test images.
Reproduction assets foundThe paper's authors explicitly state that the source code for model training and inference (including the key modules: DCMHA, eSE-enhanced FPN, Mask R-CNN/ViT pipeline) is publicly available at a GitHub repository, which matches an allowed URL. The melon stomatal image dataset (8,154 images) is described in detail but,Code · publicCode Availability
The source code for model training and inference, including the implementation of the
key modules, is publicly available at:
https://github.com/huangyao110/qk_maskrcnn_trsv2.gitOpen asset ↗huangyao110/qk_maskrcnn_trsv2pdf-page:22 lines:1-322Code / dataset availability confirmedCrossref · Europe PMC · OpenAlex · checked 5 Sept 2026
Automated quantification of plant-level development from multi-plant greenhouse scenes requires separating individual plants from shared scene-level reconstructions and quantifying organ-level development, a challenge that single-plant acquisition workflows do not directly address. This study presents an end-to-end phenotyping pipeline built on 3D Gaussian Splatting (3DGS) and a post-reconstruction extraction framework, LCR-GS, designed to isolate plant instances from full greenhouse scenes without scene-specific model retraining. LCR-GS integrates zero-shot 2D cues with multi-view lifting, geometric clustering, and chromatic refinement to convert large scene-level reconstructions (~2M Gaussians) into compact per-plant subsets (~16K Gaussians). Experiments on greenhouse-grown muskmelon at the early vegetative stage demonstrate high plant-extraction precision (0.933) and strong organ-level instance segmentation (mean AP50 = 0.924). Plant height and leaf count are validated against manual measurements (height R² = 0.98, RMSE = 1.88 cm; leaf count R² = 0.86), whereas additional morphological traits, including leaf area, leaf area index, mean internode length, and stem node count, are reported as pipeline-derived descriptors for within-cohort comparison. By decoupling semantic inference from reconstruction, the pipeline reduces scene-scale data by over 99% and provides a practical route to derive compact per-plant 3D representations from multi-plant greenhouse imagery for downstream organ-level analysis.
Why it matches plant phenotyping methods3DGS画像から個体・器官を抽出し、植物形質を定量化するフェノタイピング手法の開発と検証が中心である。
abstractThis study presents an end-to-end phenotyping pipeline built on 3D Gaussian Splatting (3DGS) and a post-reconstruction extraction framework, LCR-GS, designed to isolate plant instances from full greenhouse scenes without scene-specific model retraining.
Reproduction assets foundThe paper's Data Availability Statement points to a public GitHub repository containing the muskmelon 3DGS phenotyping dataset (scenes, Gaussian-level plant/background annotations, and point-level organ labels) used in this study.Dataset · publicThe datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://github.com/bblabNTU/3dgs-muskmelon-phenotyping-dataset.Open asset ↗bblabNTU/3dgs-muskmelon-phenotyping-datasethtml-lines:485-547Code / dataset availability confirmedEurope PMC · checked 14 Sept 2026
Wax gourd ( Benincasa hispida (Thunb.) Cogn.) and Mangalore Cucumber (Cucumis melo L. subsp. agrestis var. conomon) are nutritionally rich, mineral-dense crops with a short growing cycle, making them a preferred choice for cultivation among farmers across the country. The Mangalore cucumber, also known as the culinary cucumber, Indian yellow cucumber, or Japanese pickling melon, is widely used in Asian cuisine for pickling. While proper nutrient management is essential for optimal growth, disease control poses a significant challenge in ensuring healthy yields, as disease can rapidly spread from one leaf to another, affecting larger areas of the field and reducing crop yield. Since cucurbits grow close to the soil, they spread across the ground, exhibit dense canopies, and often overlap with neighboring plants. Early detection is crucial to ensure sustainable cultivation, food security, and increased crop productivity. To address this challenge, we collected a dataset comprising 3200 images that includes image samples of Wax gourd and Mangalore cucumber plants affected by leaf miner, pests and image samples of healthy leaves. The Cucurbitaceae datasets that are available in the public domain lack representation of the Mangalore cucumber and Wax gourd varieties. To the best of our knowledge, no publicly available dataset exists for the Wax gourd. Moreover, existing datasets typically contain images captured under controlled greenhouse conditions with plain backgrounds, featuring a single leaf per image. They exhibit low background complexity and limit the scope to detect diseases at the object level, including multiple diseases present on a single leaf or plant. The uniqueness of the proposed dataset lies in addressing this gap by providing field-level images of cucurbits. These images capture variations in soil, overlapped leaves, complex background, varying angles and distances, weeds, and human interference. This makes the dataset suitable for training object detection models capable of identifying single and multiple disease instances, and it can also be effectively used for classification tasks to distinguish between healthy and diseased leaves. It supports advancement in deep learning, feature extraction, segmentation and pattern recognition tasks. Additionally, the dataset serves as a valuable resource for plant pathologists, agronomists and agricultural experts in disease detection, monitoring and management, thereby promoting sustainable agricultural practices. By offering open access, this dataset promotes collaboration within the scientific community to facilitate the development of robust disease detection, identification, and disease control, thus enhancing farming practices and increasing agricultural yields and advancing food security.
Why it matches plant phenotyping methods植物の病害状態を画像から検出・分類する公開データセットが研究の中心であり、植物病害の画像ベース表現型解析に該当する。
abstractwe collected a dataset comprising 3200 images that includes image samples of Wax gourd and Mangalore cucumber plants affected by leaf miner, pests and image samples of healthy leaves.
Reproduction assets foundThe paper is a Data in Brief article describing the WMC-Leafset dataset of 3200 annotated field images of wax gourd and Mangalore cucumber plants for leaf miner/pest object detection. The dataset is publicly deposited on Mendeley Data with an explicit direct URL and DOI, making it a paper-specific, publicly actionable,Dataset · public6° 44′ 46″ east, latitude of 12.16999° or 12° 10′ 12″ north and Mangalore cucumber images were collected from Hulimahu village in longitude of 76.73329° or 76° 43′ 60″ east, latitude 12.1598° or 12° 9′ 35″ north
Data accessibility
Repository name: WMC-Leafset
Data identification number: 10.17632/8m2ytxd4dg.4
Direct URL to data: https://data.mendeley.com/datasets/8m2ytxd4dg/4
Related research article
[ 11 ] M. A. Keerthi Prasad, N. Shobha Rani, M. A. Sangamesha and K. V. Vinay, ``Identification and Detection of Leaf Miner, Pest Infestation in Cucurbitaceae Family in Real-Time Infield Scenarios using YOLOv5s Object Detection Model,'' 2024 11th International Conference on Computing for SustainaOpen asset ↗10.17632/8m2ytxd4dg.4lines:35-65Code / dataset availability confirmedOpenAlex · Europe PMC · checked 6 Sept 2026
Accurate fruit size estimation is crucial for plant phenotyping, as it enables precise crop management and enhances agricultural productivity by providing essential data for growth and resource efficiency analysis. In this study, we estimated the size of on-plant oriental melons grown in a vertical cultivation system to address the challenges posed by leaf occlusion. Data augmentation was achieved using a diffusion model to generate synthetic leaves to cover existing fruits and create an enriched dataset. Three instance segmentation models-mask region-based convolutional neural network (CNN), Mask2Former, and detection transformer (DETR)-and six de-occlusion models derived from these architectures were implemented. These models successfully inferred both visible and occluded areas of the fruit. Notably, Amodal Mask2Former and occlusion-aware RCNN (ORCNN) achieved average precision scores of 85.92 % and 85.35 %, respectively. The inferred masks were used to estimate the height and diameter of the fruit, with Amodal Mask2Former yielding a mean absolute error of 5.46 mm and 4.20 mm and a mean absolute percentage error of 4.86 % and 5.33 %, respectively. The results indicate enhanced performance of the transformer-based Amodal Mask2Former over CNN architectures in de-occlusion tasks and size estimation. Finally, the enhancement in de-occlusion models compared to conventional models was assessed and demonstrated across occlusion ratios ranging from 0 to 70 %. However, generating synthetic datasets with occlusion ratios over 70 % remains a limitation.
Why it matches plant phenotyping methods果実の遮蔽領域を復元し、画像から果実サイズを推定する手法の開発・評価が研究の中心であり、植物表現型計測に該当する。
abstractAccurate fruit size estimation is crucial for plant phenotyping
Reproduction assets foundThe authors explicitly state that the code for training/testing the segmentation models and the size estimation analysis is publicly available at their GitHub repository (https://github.com/sungjay-kim). The oriental melon image dataset itself is only available upon request from the corresponding author, so it is not aCode · publicThe code implemented for this study is publicly available at https://github.com/sungjay-kim . This repository contains code for training and testing segmentation models as well as algorithms for size estimation analysis.Open asset ↗https://github.com/sungjay-kimlines:553-617Code / dataset availability confirmedCrossref · Europe PMC · checked 7 Sept 2026
Abstract Background The sweet potato whitefly ( Bemisia tabaci ) is a globally important insect pest that damages crops through direct feeding and by transmitting viruses. Current B. tabaci management revolves around the use of insecticides, which are economically and environmentally costly. Host plant resistance is a sustainable option to reduce the impact of whiteflies, but progress in deploying resistance in crops has been slow. A major obstacle is the high cost and low throughput of screening plants for B. tabaci resistance. Oviposition rate is a popular metric for host plant resistance to B. tabaci because it does not require tracking insect development through the entire life cycle, but accurate quantification is still limited by difficulties in observing B. tabaci eggs, which are microscopic and translucent. The goal of our study was to improve quantification of B. tabaci eggs on several important crop species: cassava, cowpea, melon, sweet potato and tomato. Results We tested a selective staining process originally developed for leafhopper eggs: submerging the leaves in McBryde’s stain (acetic acid, ethanol, 0.2% aqueous acid Fuchsin, water; 20:19:2:1) for three days, followed by clearing under heat and pressure for 15 min in clearing solution (LGW; lactic acid, glycerol, water; 17:20:23). With a less experienced individual counting the eggs, B. tabaci egg counts increased after staining across all five crops. With a more experienced counter, egg counts increased after staining on melons, tomatoes, and cowpeas. For all five crops, there was significantly greater agreement on egg counts across the two counting individuals after the staining process. The staining method worked particularly well on melon, where egg counts universally increased after staining for both counting individuals. Conclusions Selective staining aids visualization of B. tabaci eggs across multiple crop plants, particularly species where leaf morphological features obscure eggs, such as melons and tomatoes. This method is broadly applicable to research questions requiring accurate quantification of B. tabaci eggs, including phenotyping for B. tabaci resistance.
Why it matches plant phenotyping methods植物葉上のコナジラミ卵を染色して定量し、計数値と計数者間一致を改善する方法を評価しており、抵抗性フェノタイピングへの応用が明示された中心的な手法研究。
abstractThe goal of our study was to improve quantification of B. tabaci eggs on several important crop species: cassava, cowpea, melon, sweet potato and tomato.
Reproduction assets foundThe paper's Data availability statement explicitly deposits the study's egg-count datasets and the R Markdown analysis code in a Dryad repository, which is a public, paper-specific asset directly reproducing the phenotyping measurements and analysis.Dataset · publicThe datasets generated and analyzed during this study, and an R Markdown document containing the code used to perform these analyses are available in a Dryad repository (DOI: doi: https://doi.org/10.5061/dryad.vmcvdnd1m ).Open asset ↗Dryad · 10.5061/dryad.vmcvdnd1mlines:138-163Code / dataset availability confirmedOpenAlex · Europe PMC · Crossref · checked 14 Sept 2026
Abstract Cucumis melo L., commonly known as melon, is a crucial horticultural crop. The selection and breeding of superior melon germplasm resources play a pivotal role in enhancing its marketability. However, current methods for melon appearance phenotypic analysis rely primarily on expert judgment and intricate manual measurements, which are not only inefficient but also costly. Therefore, to expedite the breeding process of melon, we analyzed the images of 117 melon varieties from two annual years utilizing artificial intelligence (AI) technology. By integrating the semantic segmentation model Dual Attention Network (DANet), the object detection model RTMDet, the keypoint detection model RTMPose, and the Mobile-Friendly Segment Anything Model (MobileSAM), a deep learning algorithm framework was constructed, capable of efficiently and accurately segmenting melon fruit and pedicel. On this basis, a series of feature extraction algorithms were designed, successfully obtaining 11 phenotypic traits of melon. Linear fitting verification results of selected traits demonstrated a high correlation between the algorithm-predicted values and manually measured true values, thereby validating the feasibility and accuracy of the algorithm. Moreover, cluster analysis using all traits revealed a high consistency between the classification results and genotypes. Finally, a user-friendly software was developed to achieve rapid and automatic acquisition of melon phenotypes, providing an efficient and robust tool for melon breeding, as well as facilitating in-depth research into the correlation between melon genotypes and phenotypes.
Why it matches plant phenotyping methodsメロン果実・果梗を画像から分割し、特徴抽出によって11形質を推定する深層学習フレームワークを開発・検証し、ソフトウェア化しているため、植物フェノタイピング手法が中心である。
abstractBy integrating the semantic segmentation model Dual Attention Network (DANet), the object detection model RTMDet, the keypoint detection model RTMPose, and the Mobile-Friendly Segment Anything Model (MobileSAM), a deep learning algorithm framework was constructed, capable of efficiently and accurately segmenting melon fruit and pedicel.
Reproduction assets found保存済みの本文根拠を更新済みルールで再検証し、公開資産1件を確認しました。Code · publicAdditionally, we have developed a simple melon phenotypic traits extraction software, which can be downloaded via https://github.com/hongbinz13/Melon-Phenotype-Extractor/releases/tag/software .Open asset ↗https://github.com/hongbinz13/Melon-Phenotype-Extractor · Melon-Phenotype-Extractorlines:109-139Code / dataset availability confirmedEurope PMC · Crossref · checked 15 Sept 2026
Flower morphologies shape the accessibility to nectar and pollen, two major traits that determine plant-pollinator interactions and reproductive success. Melon is an economically important crop whose reproduction is completely pollinator-dependent and, as such, is a valuable model for studying crop-ecological functions. High-resolution imaging techniques, such as micro-computed tomography (micro-CT), have recently become popular for phenotyping in plant science. Here, we implemented micro-CT to study floral morphology and honey bees in the context of nectar-related traits without a sample preparation to improve the phenotyping precision and quality. We generated high-quality 3D models of melon male and female flowers and compared the geometric measures. Micro-CT allowed for a relatively easy and rapid generation of 3D volumetric data on nectar, nectary, flower, and honey bee body sizes. A comparative analysis of male and female flowers showed a strong positive correlation between the nectar gland volume and the volume of the secreted nectar. We modeled the nectar level inside the flower and reconstructed a 3D model of the accessibility by honey bees. By combining data on flower morphology, the honey bee size and nectar volume, this protocol can be used to assess the flower accessibility to pollinators in a high resolution, and can readily carry out genotypes comparative analysis to identify nectar-pollination-related traits.
Why it matches plant phenotyping methodsマイクロCTを用いて花、蜜腺、蜜、ハナバチの3D形態・体積を取得し、花粉媒介関連形質を高精度に評価するプロトコルを実装・提示しており、表現型取得法が中心である。
abstractHere, we implemented micro-CT to study floral morphology and honey bees in the context of nectar-related traits without a sample preparation to improve the phenotyping precision and quality.
Reproduction assets foundThe paper deposits its Python image-processing/phenotyping pipeline on GitHub and provides a supplement containing raw nectar/nectary measurement data (Tables S1–S4). Both are paper-specific, public, and actionable.Supplement · publicThe following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/cells11213452/s1 . Figure S1: pollen on Stamens in ♂ and ⚥ flower types at different magnifications; Table S1: nectar-related traits in male and female flowers; Table S2: correlation analysis between nectary volume, nectary cross-section area, nectary surface, flower width and nectar volume in the respective male, female and pooled melon flowerOpen asset ↗lines:83-224Code / dataset availability confirmedEurope PMC · bioRxiv · checked 15 Sept 2026
The generation of realistic plant and animal images from marker information could be a main contribution of artificial intelligence to genetics and breeding. Since morphological traits are highly variable and highly heritable, this must be possible. However, a suitable algorithm has not been proposed yet. This paper is a proof of concept demonstrating the feasibility of this proposal using ‘decoders’, a class of deep learning architecture. We apply it to Cucurbitaceae, perhaps the family harboring the largest variability in fruit shape in the plant kingdom, and to tomato, a species with high morphological diversity also. We generate Cucurbitaceae shapes assuming a hypothetical, but plausible, evolutive path along observed fruit shapes of C. melo . In tomato, we used 353 images from 129 crosses between 25 maternal and 7 paternal lines for which genotype data were available. In both instances, a simple decoder was able to recover expected shapes with large accuracy. For the tomato pedigree, we also show that the algorithm can be trained to generate offspring images from their parents’ shapes, bypassing genotype information. Data and code are available at https://github.com/miguelperezenciso/dna2image .
Why it matches plant phenotyping methodsDNA配列や親の形状から植物果実形状画像を生成する深層学習手法の概念実証であり、植物形態の取得・推定が研究の中心です。
titleComputer generation of fruit shapes from DNA sequence
Reproduction assets foundThe paper's cucurbit shape phenotyping inputs and analysis code are publicly available in the authors' dna2image GitHub repository, explicitly cited in the methods and data availability statement.Dataset · publichways. One pathway would be wild gourd (akin to pumpkin shape) scallop acorn; a
134
second pathway would be wild gourd marrow straightneck zucchini cocozelle
135
(Figure 1B). See also Figure 17 in (Paris 1989). We extracted contours from the
136
‘contours.png’ file, based in (Paris 1989) and available in GitHub
137
(https://github.com/miguelperezenciso/dna2image/blob/main/images/contours.png), using
138
OpenCV library (Bradski 2000). Contours were centered and 500 pseudo-landmarks were
139
obtained with the algorithm in Zingaretti et al. (2021). Next, contours were aligned with a
140
generalized procrustes algorithm implemented in python package ‘procrustes’ (Meng et al.
141
2022Open asset ↗https://github.com/miguelperezenciso/dna2image · contours.pngpdf-raw-page:5 lines:1-76Code · publicy, we have shown that very simple networks can be successfully trained in small
322
datasets to accurately predict fruit images. Although much work remains to be done, this
323
research opens new possibilities in the area of prediction of complex traits.
324
325
Data availability statement
326
All data and code are available at https://github.com/miguelperezenciso/dna2image.327
328
.
CC-BY 4.0 International license
available under a
(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made
The copyright holder for this preprint
this version posted September 22, 2022.
;
https://doi.org/10.1101/2022.09.19.Open asset ↗https://github.com/miguelperezenciso/dna2image.327 · dna2image.327pdf-raw-page:10 lines:1-73