Farmers must be able to estimate their crop yields at various growth stages for effective management of their farms and to enable them to interact with cooperatives or traders as early as possible. Here we developed an AI-based cocoa pod detection method using low resolution colour images of cocoa trees on farms in Côte d'Ivoire. We compared nano and extra-large architectures of six neural networks, including Faster RCNN, Baidu's Real-Time Detection Transformer (RTDetr), Detr-ResNet Vision Transformer (ViT), YOLOv5, YOLOv8 and YOLOv11. These networks were trained with 7,850 annotated cocoa pods on 400 low resolution images, and validated in two independent datasets: a 42 low resolution images containing 990 annotated pods, and a 100 low resolution images containing 2,400 annotated pods. The performances of the nano YOLOv8 and YOLOv11 networks were 2% higher than that of the RTDetr networks and 5% higher than that of the YOLOv5, ViT and Faster RCNN networks with an F1-score of 77% on all images and up to 90% on foreground trees. The dominance of nano architectures suggests that the extra-large architectures, which contain 20-30-times more neurons, may not have been fully trained. The study of learning performance curves showed that extra-large networks were unable to outperform nano networks, which contradicts the theory. After review, the annotated dataset was found to contain inconsistencies. The inconsistency of the training and validation data and their limited quantity restricted the objectivity of comparisons between network architectures. Finally, although the average detection performance of RTDetr for cocoa pods was only 2% lower than that of the YOLOv8 network, it was definitively excluded from the candidate models because its per-image processing time was 15-20% higher than that of YOLOv8 and YOLOv11. However, with a performance sensitivity to data of less than 0.5%, YOLOv8 Nano became the best option.
Why it matches plant phenotyping methodsカカオ果実を画像から検出・定量するAI手法を開発し、複数モデルと独立データセットで性能比較・検証しており、植物フェノタイピング手法が中心である。
abstractHere we developed an AI-based cocoa pod detection method using low resolution colour images of cocoa trees on farms in Côte d'Ivoire.
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。Dataset · publicData Availability: The data used in the study can be downloaded from CIRAD’s data verse at https://doi.org/10.18167/DVN1/8COJBB .Open asset ↗10.18167/DVN1/8COJBBlines:129-140Code / dataset availability confirmedEurope PMC · checked 5 Sept 2026
Accurate detection of cocoa pod diseases is vital to reducing yield losses and supporting sustainable agriculture. Although deep learning models have shown promise in plant disease classification, their performance often varies between datasets due to limitations in feature extraction and generalisation. This study introduces a Learnable Gated Fusion Convolutional Block Attention Module (LGF-CBAM) integrated with a ResNetV2-101 backbone to improve discriminative feature learning and improve robustness in cocoa disease classification. Unlike the standard CBAM, which processes attention modules sequentially, LGF-CBAM adaptively balances the importance of spatial and channel cues through trainable gating parameters normalized with a softmax function. Incorporating LGF-CBAM provided outstanding results on the Cocoa_Pod_Disease_Gh dataset, achieving 98.95% accuracy along with F1 and PPV scores of 99.11%. The cross-dataset evaluation confirmed robustness, with accuracies of 98.53% on Cocoa Diseases (YOLOv4), 97.96% on Black and Borer Pod Rot, and 96.19% on Cacao Diseases in Davao. Although greater variability in the Coffee and Cocoa dataset reduced accuracy to 94.00%, the model still maintained strong adaptability under diverse conditions. These findings establish LGF-CBAM as a state-of-the-art framework that outperforms all other referenced systems, offering high accuracy, stability, and generalization. In general, this research contributes to a novel attention-based deep learning framework that can support early and reliable identification of cocoa pod diseases, providing a scalable solution for precision agriculture.
Why it matches plant phenotyping methodsカカオ果実の病害状態を画像から分類する深層学習手法を開発し、複数データセットで性能・頑健性を評価しており、植物フェノタイピング手法が研究の中心である。
abstractThis study introduces a Learnable Gated Fusion Convolutional Block Attention Module (LGF-CBAM) integrated with a ResNetV2-101 backbone to improve discriminative feature learning and improve robustness in cocoa disease classification.
Reproduction assets foundThe authors' primary plant-phenotyping asset is the Cocoa_Pod_Disease_Gh image dataset, publicly deposited on Figshare with an explicit Data Availability statement and DOI. The Kaggle/Roboflow datasets are cited prior external datasets used for cross-dataset evaluation, not paper-specific deposits, so they are excludedDataset · publicThe data that support the findings of this study is available at https://figshare.com/articles/dataset/Cocoa_Disease_Datasets/31294003. https://doi.org/10.6084/m9.figshare.31294003.Open asset ↗figshare · 10.6084/m9.figshare.31294003html-lines:1039-1062Code / dataset availability confirmedEurope PMC · checked 6 Sept 2026
Premise This study investigates advanced training techniques to improve the performance of convolutional neural networks for disease detection in cocoa, Theobroma cacao . Methods Despite recent stagnation in accuracy improvements in computer vision for image classification, our research demonstrates significant advancements in performance through semi-supervised learning, specialised loss functions, and the inclusion of a non-cocoa class. Results Semi-supervised learning reduced overfitting and enhanced generalisability, particularly for subtle symptoms. The non-cocoa class exposed models to a broad range of relevant features, significantly improving model robustness and performance in difficult cases. Grad-CAM for qualitative assessment provided valuable insights into model behaviour, highlighting cases of overfitting missed by summary statistics. We also describe dynamic focal loss, a novel loss function that uses an empirical measure of difficulty to weight each image. Our results suggest that while PhytNet shows promise in terms of computational efficiency and superior handling of difficult images, ResNet18 with semi-supervised learning and dynamic focal loss emerged as the strongest contender for real-world deployment. Discussion This research underscores the potential of semi-supervised learning and advanced loss functions in enhancing the applicability of deep learning models in agricultural disease management. It also presents a new high-quality benchmark dataset of 7220 images of diseased and healthy cocoa trees, offering a much greater and more realistic challenge than the Plan Village dataset.
Why it matches plant phenotyping methodsカカオ葉・樹体の病徴画像から植物の病害状態を推定する深層学習手法を開発・比較し、性能評価とベンチマークデータセット構築を行っており、フェノタイピング手法が中心である。
abstractThis study investigates advanced training techniques to improve the performance of convolutional neural networks for disease detection in cocoa, Theobroma cacao .
Reproduction assets foundThe paper's data availability statement explicitly provides the paper-specific cocoa image dataset and the FAIGB dataset on OSF, plus authors' analysis code on GitHub, all with public URLs.Dataset · publicThe cocoa image data is available at https://osf.io/2fw6gOpen asset ↗osf · 2fw6glines:753-923Dataset · publicthe FAIGB web‐scraped dataset of crop disease images is available at https://osf.io/nuafhOpen asset ↗osf · nuafhlines:753-923Code · publicAll code necessary to reproduce these results is available on GitHub ( https://github.com/jrsykes/CocoaReader/tree/main/CocoaNet/PhytNet_Cocoa )Open asset ↗github · jrsykes/CocoaReaderlines:753-923Code / dataset availability confirmedEurope PMC · checked 7 Sept 2026
Premise Automated disease, weed, and crop classification with computer vision will be invaluable in the future of agriculture. However, existing model architectures like ResNet, EfficientNet, and ConvNeXt often underperform on smaller, specialised datasets typical of such projects. Methods We address this gap with informed data collection and the development of a new convolutional neural network architecture, PhytNet. Utilising a novel dataset of infrared cocoa tree images, we demonstrate PhytNet's development and compare its performance with existing architectures. Data collection was informed by spectroscopy data, which provided useful insights into the spectral characteristics of cocoa trees. Cocoa was chosen as a focal species due to the diverse pathology of its diseases, which pose significant challenges for detection. Results ResNet18 showed some signs of overfitting, while EfficientNet variants showed distinct signs of overfitting. By contrast, PhytNet displayed excellent attention to relevant features, almost no overfitting, and an exceptionally low computation cost of 1.19 GFLOPS. Conclusions We show that PhytNet is a promising candidate for rapid disease or plant classification and for precise localisation of disease symptoms for autonomous systems. We also show that the most informative light spectra for detecting cocoa disease are outside the visible spectrum and that efforts to detect disease in cocoa should be focused on local symptoms, rather than the systemic effects of disease.
Why it matches plant phenotyping methods植物病害画像から症状を検出・局在化するCNNアーキテクチャPhytNetを開発し、既存モデルと比較検証しており、植物表現型取得・抽出法が中心である。
abstractthe development of a new convolutional neural network architecture, PhytNet
Reproduction assets foundThe paper's cocoa disease image/spectroscopy data are deposited on OSF (freely accessible via the provided link) and the PhytNet training/optimisation code is publicly available on GitHub. Both are paper-specific, public, and actionable.Code · publicThe code to optimise and train PhytNet for your data can be found at: https://Github.com/jrsykes/PhytNet .Open asset ↗Github · jrsykes/PhytNetlines:214-297Code / dataset availability confirmedCrossref · Europe PMC · checked 13 Sept 2026
Abstract Plant pathogens can decimate crops and render the local cultivation of a species unprofitable. In extreme cases this has caused famine and economic collapse. Timing is vital in treating crop diseases, and the use of computer vision for precise disease detection and timing of pesticide application is gaining popularity. Computer vision can reduce labour costs, prevent misdiagnosis of disease, and prevent misapplication of pesticides. Pesticide misapplication is both financially costly and can exacerbate pesticide resistance and pollution. Here, we review the application and development of computer vision and machine learning methods for the detection of plant disease. This review goes beyond the scope of previous works to discuss important technical concepts and considerations when applying computer vision to plant pathology. We present new case studies on adapting standard computer vision methods and review techniques for acquiring training data, the use of diagnostic tools from biology, and the inspection of informative features. In addition to an in‐depth discussion of convolutional neural networks (CNNs) and transformers, we also highlight the strengths of methods such as support vector machines and evolved neural networks. We discuss the benefits of carefully curating training data and consider situations where less computationally expensive techniques are advantageous. This includes a comparison of popular model architectures and a guide to their implementation.
Why it matches plant phenotyping methods植物病害を対象としたコンピュータビジョンによる症状・病害の検出手法を中心に扱うレビューであり、植物フェノタイピング手法の方法論的整理と評価が主題である。
abstractHere, we review the application and development of computer vision and machine learning methods for the detection of plant disease.
Reproduction assets foundThe paper's data availability statement provides public OSF deposits (view-only links) containing image data, annotations, training data, and semi-supervised model weights for the cocoa disease-detection case studies, plus public GitHub repositories with the authors' custom training/analysis code (CocoaReader, CocoaNetDataset · publicThe image data, annotations, and link to the accompanying GitHub repository for Case Study 1 can be found at: https://osf.io/79kx3/?view_only=4a2c1dccee1a4baeb85de5002c702f10 .Open asset ↗osflines:411-466Dataset · publicFor Case Study 2, the data used to train the initial supervised model, the .csv search terms file for the below web scraper, and the final semi‐supervised model weights can be found at: https://osf.io/h5gj7/?view_only=dbf9f245e21a41e185f5b73e718b4cad .Open asset ↗osflines:411-466Code · publicThe custom code used to train both the initial model and the final semi‐supervised model can be found at: https://github.com/jrsykes/CocoaReader/blob/main/PlantNotPlant .Open asset ↗github · jrsykes/CocoaReaderlines:411-466Code · publicThe custom code to run the sweep in Case Study 4 can be found in the following GitHub repository: https://github.com/jrsykes/CocoaReader/tree/main/CocoaNet .Open asset ↗github · jrsykes/CocoaReaderlines:411-466Dataset · publicThe data used to generate these results and the full wandb report can be found at: https://osf.io/2fw6g/?view_only=adc66ba66f83465a9e7b111515a60bf2 .Open asset ↗osflines:411-466Code · publicThe “contaminated” data used to train the semi‐supervised model were generated using the code at: https://github.com/jrsykes/Google-Image-Scraper .Open asset ↗github · jrsykes/Google-Image-Scraperlines:411-466Code / dataset availability confirmedEurope PMC · checked 14 Sept 2026
Cocoa cultivation is the basis for chocolate production; it has a unique aroma that makes it useful in the production of snacks and usable for cooking or baking. The maximum harvest period of cocoa is normally once or twice a year and spread over several months, depending on the country. Determining the best harvesting period for cocoa pods plays a major role in the export process and the pods quality. The degree of ripening of the pods affects the quality of the resulting beans. Also, unripe pods do not have enough sugar and may prevent proper bean fermentation. As for too-mature pods, they are usually dry, and their beans may germinate inside the pods, or they may develop a fungal disease and cannot be used. Computer-based determination of the ripeness of cocoa pods throughout image analysis could facilitate massive cocoa ripeness detection. Recent technological advances in computing power, communication systems, and machine learning techniques provide opportunities for agricultural engineering and computer scientists to meet the demands of the manual. The need for diverse and representative sets of pod images is essential for developing and testing automatic cocoa pod maturity detection systems. In this perspective, we collected images of cocoa pods to set up a database of cocoa pods of the Côte d'Ivoire named CocoaMFDB. We performed a pre-processing step using the CLAHE algorithm to improve the quality of the images since the effect of the light was not controlled on our data set. CocoaMFDB allows the characterization of cocoa pods according to their maturity level and provides information on the pod family for each image. Our dataset comprises three large families, namely Amelonado, Angoleta, and Guiana, grouped into two maturity categories: the ripe and unripe pods. It is, therefore, perfect for developing and evaluating image analysis algorithms for future research.
Why it matches plant phenotyping methodsカカオ果実の成熟度という植物器官形質を対象とする画像データセットを構築し、画像解析アルゴリズムの開発・評価用に提供しているため、フェノタイピング手法・データセットが中心です。
abstractwe collected images of cocoa pods to set up a database of cocoa pods of the Côte d'Ivoire named CocoaMFDB.
Reproduction assets foundThe paper is a data descriptor for CocoaMFDB, a public dataset of cocoa pod images (maturity/family) with PASCAL VOC XML annotations, deposited on Mendeley Data with a direct URL and DOI matching an allowed URL.Dataset · publiced pods.
Data source location
The images of cocoa pods are from the plantations of Yakassé 1, a village of Grand Bassam first capital of Côte d'Ivoire with a Latitude and longitude of 5°12′42″ north, 3°44′19″.
Data accessibility
Repository name: Mendeley Data
Data identification number: 10.17632/9msjjh3np6.2
Direct URL to data: https://data.mendeley.com/datasets/9msjjh3np6/2
Value of the Data
•
Images of cocoa pods obtained will be used in identifying cocoa types and varieties.
•Open asset ↗Mendeley Data · 10.17632/9msjjh3np6.2lines:1-52