AbstractIn Burkina Faso, the mango industry faces numerous phytosanitary constraints. This situation is exacerbated by producers’ lack of knowledge about diseases and their symptoms and the low uptake of control technologies. This study aims to develop a tool using artificial intelligence-based approaches for better integrated management of the main diseases affecting mango trees. Surveys conducted in 40 mango orchards yielded 11,001 images of the characteristic symptoms of three major diseases, including 4,527 for anthracnose caused by Colletotrichum gloeosporioides, 3,038 for bacterial disease caused by Xanthomonas citri pv. mangiferaeindicae and 3,436 for dieback caused by Lasiodiplodia spp. The data underwent several preprocessing steps. To ensure a balance between disease classes, 3,000 images per class were selected, for a total of 9,000 images. The images were then annotated and used for training with the pre-trained YOLOv11 model. Following training, three models were formed : V2-3 M (MAP@50 = 50,3%; precision = 55,1%; recall = 49,1%; F1-score = 51,9%), V3-3 M (MAP@50 = 23%; precision = 56,8% ; recall = 26,7%; F1-score = 36,3%), and V4-3 M (MAP@50 = 46,8%; precision = 52%; recall = 45,3%; F1-score = 48,4%). The most effective model is the V2-3 M, as it combines 55,1% accuracy, 50,3% (MAP@50, 49,1% recall and F1-score = 51,9%). In order to improve diagnosis and phytosanitary management in orchards, it would be useful to integrate treatment recommendations into the model, then deploy it on a mobile application and promote it to producers. Future work will involve dataset expansion, comparative evaluation with alternative deep learning models, and cross-regional validation to improve model robustness and generalizability.IntroductionMangoes are the fifth most produced fruit in the world (by tonnage) behind bananas/plantains, citrus fruits, grapes, and apples, with a production of 51 million tonnes in 20171. Burkina Faso contributes between 11% and 18% of mango production in West Africa2. Mangoes are a very important socio-economic and climate issue in Burkina Faso. Mango production and export contribute significantly to the country’s income.Indeed, the mango sector generates more than 15 billion in turnover per year1. In addition, it contributes to food security in producer areas and vegetation cover, limiting the effects of climate change2. The mango value chain in Burkina Faso is a source of employment. Its activities employ 28,000 people in 21,000 jobs as entrepreneurs or similar (farmers, trackers, retailers, etc.)2, around 350 permanent jobs in processing and packaging companies for fresh exports, and around 7,000 seasonal jobs throughout the chain3.The mango sector faces enormous phytosanitary constraints, particularly insect pests4 and also major diseases such as anthracnose caused by Colletotrichum gloeosporioides, bacterial black spot disease caused by Xanthomonas citri pv. mangiferaeindicae and dieback5,6,7 caused by Lasiodiplodia spp. These diseases cause significant damage in orchards, such as deterioration in fruit quality, tree death and reduced yields8,9. This damage causes enormous losses, such as loss of income for producers, management costs, reduced productivity, etc.10 Unfortunately, producers have little understanding of these diseases and their symptoms, epidemiological aspects and diagnostic difficulties. Moreover, the limited dissemination of available control technologies, resulting in their low adoption by producers. Nowadays, artificial intelligence is a powerful decision-making tool in agriculture, particularly in the automatic diagnosis of crop diseases11. In India, automatic diagnosis and classification models for crop pests have been implemented12,13.This aspect of AI for agricultural development is still largely unexplored in Burkina Faso, specifically for the diagnosis of mango tree diseases. The aim of this study is to design an effective AI-based automatic diagnosis model for the main mango tree diseases in Burkina Faso.Materials and methodsStudy areaThe study was conducted in the Guiriko (Houet and Kénédougou provinces) and Tannounyan (Comoé and Léraba provinces) regions of Burkina Faso (Fig. 1), which account for approximately 70% of national mango production14. The Guiriko region has a Sudano-Guinean tropical climate with annual rainfall ranging from 900 to 1200 mm, while the Tannounyan region receives between 1200 and 1700 mm per year15, creating agroclimatic conditions highly favorable to mango production. Mango orchards in this area are affected by major diseases such as dieback7, bacterial disease6 and anthracnose5. The local economy is based on agriculture, livestock farming and trade, with the mango industry supported by industrial and artisanal processing units7,16.Fig. 1Full size imageMap showing the location of sites in the study area.Collection of image data on diseases in mango orchardsSurveys, collection, and compilation of representative imagesMango orchards were selected from the membership list provided by the Mangoes Producers Association of Burkina (APROMAB), based on orchard accessibility and producer consent. Field surveys and data collection were carried out from March to July 2024, corresponding to the main mango production and disease expression period. Ten (10) mango orchards were selected per province, for a total of 40 orchards across the four (04) provinces (Kénédougou, Houet, Comoé and Leraba). These same orchards were used for both disease incidence and severity assessments and image data collection. A total of 2,000 mango trees (50 trees per orchard) were inspected. Mango trees were systematically inspected along diagonal transects. Consecutive sampled trees were separated by an average distance of approximately 10 m, corresponding to the regular tree spacing within the orchards.This operation was conducted by a multidisciplinary team of specialists, including plant pathologists, an entomologist, and a computer scientist, ensuring both phytopathological expertise and technical rigor in image acquisition.Within each orchard, mango trees showing symptoms of the three main diseases including anthracnose caused by Colletotrichum gloeosporioides, bacterial black spot disease caused by Xanthomonas campestris pv. mangiferaeindicae and dieback caused by Lasiodiplodia spp. were identified through systematic visual inspection along orchard diagonals. Disease severity was assessed using standardized rating scales: scale for anthracnose,17 scale for bacterial disease18, and scale for dieback, each ranging from 1 (no symptoms) to 5 (severe infection). Disease incidence (Ti) and severity (Ts) were calculated using the following formulas:$$ {\text{Incidence:}}\quad {\text{Ti}} = \left( {{\text{n}}/{\text{N}}} \right) \times {{1}}00 $$where n is the number of diseased trees per province and N is the total number of trees observed (N = 100).$$ {\text{Severity:}}\quad {\text{Ts}} = [\Sigma ({\text{Xi}} \times {\text{ni}})/({\text{N}} \times {\text{Z}})] \times {\text{1}}00 $$where Xi is the severity score of tree i, ni is the number of trees with severity score i, N is the total number of trees observed (N = 100), and Z is the maximum scale value (Z = 5).Photographs of disease symptoms were taken using Android V12 smartphones and a Camon 3 EOS camera between 8am and 4pm (daytime) under natural field conditions and included variations in illumination, symptom severity, viewing angle, and background complexity to better reflect real-world production environments. Symptoms were captured on leaves, fruits, and whole trees in cases of severe dieback. Images were stored on an external hard drive (1 TB), organized by disease class. The complete image dataset is publicly available at https://www.ineraopendata.com.Pre-processing and image annotationThe images were pre-processed to improve their quality and usability. This involved resizing them to dimensions compatible with YOLOv1119, normalizing colors, and correcting brightness. From the 11,001 images collected in the field, 3000 images were selected per disease class (anthracnose: 3000/4527; bacterial disease: 3000/3038; dieback: 3000/3436), for a total of 9000 images, ensuring a balanced class distribution. Image selection followed a stratified random sampling procedure, ensuring representation of all major symptom manifestations of each disease. Near-duplicate images defined as multiple shots of the same organ or lesion differing only in camera angle or distance were excluded to avoid redundancy and potential overfitting. These images were imported into the Roboflow platform (https://app.roboflow.com/) for annotation. Annotation was carried out from July to November 2024 by a multidisciplinary team comprising plant pathologists, an entomologist, and a data scientist. Image batches were distributed among annotators, each responsible for drawing bounding boxes around visible disease symptoms and assigning the corresponding disease class label (Fig. 2).Fig. 2Full size imageAnnotation operation on the Robboflow platform.Following this first annotation pass, all annotations underwent systematic quality control review: each annotated image was re-examined to verify the accuracy and consistency of bounding boxes and class assignments. Any incorrectly annotated images were identified and corrected before being included in the final dataset. This two-step process, annotation followed by expert validation, ensured the reliability of the labeled dataset used for model training. Data augmentation was applied to improve model robustness, using the following specific transformations: horizontal flipping, 90° clockwise and counter-clockwise rotation, saturation adjustment (between − 34% and + 34%), Gaussian blur (up to 1 px), and noise addition (up to 1.97% of pixels). These augmentation parameters were chosen to simulate realistic variations in field imaging conditions while preserving the visual integrity of disease symptoms.The annotation process generated a structured and standardized dataset suitable for supervised object detection tasks.Image trainingSupervised transfer learning with the pre-trained YOLOv11 algorithm was used. YOLOv11 (You Only Look Once, version 11)19 is a convolutional neural network (CNN) designed for real-time object detection, incorporating key architectural enhancements for improved performance19. The entire supervised training process was conducted via the Roboflow platform (Roboflow, 2024), which provides an integrated environment for managing datasets and training computer vision models.The annotated dataset of 9,000 images was first divided into three stratified subsets prior to any augmentation: 60% for training (5,400 images), 30% for validation (2,700 images), and 10% for testing (900 images). Splitting was stratified by disease class, ensuring that each subset contained a proportionally representative distribution of anthracnose, bacterial disease, and dieback images. This stratified random splitting was performed before augmentation to strictly prevent data leakage between subsets.Data augmentation was subsequently applied exclusively to the training set, expanding it from 5,400 to 10,800 images. The final complete dataset thus comprised 14,400 images in total (10,800 augmented training images + 2,700 validation images + 900 test images). The validation and test sets retained only original, unmodified images, ensuring that performance metrics reflect genuine generalization ability under real field conditions.Regularization techniques such as early stopping was applied to avoid overfitting.To assess the generalization capability of the best-performing model (V2-3 M) beyond the internal test set, an inference evaluation was conducted on an independent set of 115 images collected from additional mango trees in the Bobo-Dioulasso locality, not included in the original 40 mango orchards used for training. At a confidence threshold of 70%, the model correctly detected disease symptoms in 104 out of 115 images, yielding an inference accuracy of 90.4%. This result suggests that the V2-3 M model demonstrates promising generalization ability under real field conditions beyond its training environment.Models performance was evaluated using standard object detection metrics such as :Precision : the proportion of correct predictions among all predictions made. It indicates the model’s ability to limit false detections.Recall : proportion of cases correctly detected in relation to all existing positive cases. It reflects the model’s ability to not omit any symptoms that are present.mAP@50 (mean Average Precision at 50% IoU) : an overall indicator of object detection performance, calculated as the average precision across all classes, with an intersection over union (IoU) threshold set at 50%.F1-score : a complementary evaluation metric to assess the balance between precision and recall.Data processing and analysisThe data was organized and classified in Excel 2019. Excel was used to total the number of images per disease and to create graphs. The GIS tool (My Coordinates) was used to record the geographical coordinates of the various sites for mapping purposes using QGIS software. Robboflow software was used for image annotation and model training.Prior to model comparison, a non-parametric Kruskal-Wallis test was performed to assess whether significant differences existed among the evaluated models. The results of the statistical analysis, combined with the comparative evaluation of performance metrics, were used to identify the best-performing model. Statistical analyses were conducted using Python 3.13 (Python Software Foundation, Wilmington, DE, USA) with the SciPy statistical library. Statistical significance was assessed at α = 0.05. In addition, descriptive and comparative analyses were performed to identify the model showing the best overall detection performance.ResultsSymptoms of the three main annotated diseasesSurveys carried out in the four provinces enabled the symptoms of anthracnose, bacterial disease, and dieback to be identified and annotated using the Roboflow platform.Anthracnose is characterized by black spots that develop into necrosis on the leaves and deep necrotic lesions on the fruit (Fig. 3).Fig. 3Full size imageAnnotated symptoms of anthracnose on mango organs. (A) and (B): on leaves; (C): on fruit.Bacterial disease manifests as irregular oily spots on the leaves, sometimes perforated, and blackish star-shaped and cracked lesions on the fruit, sometimes with bacterial exudate (Fig. 4). The dieback caused the leaves to wilt and gradually turn brown, with partial (10 to 90%) and total (100%) coverage (Fig. 5).Fig. 4Full size imageAnnotated symptoms of bacterial blight on mango organs. (A): on leaves (B) and (C): on fruit.Fig. 5Full size imageAnnotated symptoms of dieback on mango organs. (A): on leaves, (B): partial dieback (C): total dieback.Performance of automatic detection models for major diseasesTraining identified three model variants named V2-3 M, V3-3 M, and V4-3 M, trained with 71, 38, and 75 epochs, respectively. These models were evaluated based on standard metrics, namely precision, recall, F1-score, and Mean Average Precision at an IoU of 50% (mAP@50) (Fig. 6). The evaluation reveals contrasting performances in the detection of symptoms of anthracnose, bacterial disease and dieback. The V2-3 M model stands out with the best overall results (mAP@50 = 50.3%; precision = 55.1%; recall = 49.1%; F1-score = 51.9%) and consistent per-class detection (F1: anthracnose = 14.2%; bacterial disease = 41.5%; dieback = 42.6%). The V4-3 M model, despite being trained with the highest number of epochs (75), performs slightly worse (mAP@50 = 46.8%; precision = 52.0%; recall = 45.3% ; F1 = 48.4%), while V3-3 M, trained with only 38 epochs, shows the weakest results (mAP@50 = 23.5%; precision = 56.8%; recall = 26.7%; F1-score = 36.3%), including a complete failure to detect dieback (precision = recall = F1 = 0.0%). To further compare the trained models, a Kruskal–Wallis test was performed using Precision, Recall and F1-score. No statistically significant differences were observed among the models for Precision (H = 2.222, P = 0.329), Recall (H = 3.854, P = 0.146) or F1-score (H = 3.822, P = 0.148). Nevertheless, descriptive analysis consistently identified the V2-3 M model as the best-performing model, achieving the highest overall values for mAP@50, Precision, Recall and F1-score compared to V4-3 M and V3-3 M.Fig. 6Full size imageModel characteristics after symptom training.DiscussionDistinctive symptoms of mango tree diseases : a solid basis for establishing an automatic diagnostic systemAccurate identification of phytopathological symptoms is a fundamental prerequisite for the development of digital diagnostic tools, particularly those using AI. The symptoms documented in mango orchards largely corroborate the classic descriptions established by Scientists. Indeed, for anthracnose, observations of irregular brown spots on leaves, sunken black lesions on fruit, and necrosis of young twigs are consistent with the work of 20,21, which highlighted similar symptoms of anthracnose in humid mango-producing areas, where relative humidity and poor aeration promote infection. The work of 22 has also demonstrated the early appearance of symptoms as early as the flowering stage, directly influencing yield.With regard to bacterial disease, the photographed symptoms captured, such as translucent oily spots, leaf perforations, and exudates on fruit, are typical of infection by Xanthomonas campestris pv. mangiferaeindicae. These symptoms were described by 6 in Burkina Faso and West Africa, who highlighted the rapid progression of bacterial lesions. Recognition of these visual characteristics is crucial in order to differentiate this disease from anthracnose, whose lesions can sometimes overlap on the fruit. With regard to decline, symptomatic analysis reveals leaf wilting, progressive browning of twigs, and fruit abortion7. This description corresponds to the symptoms associated with complex fungal infections, particularly Lasiodiplodia spp., as reported by23.All of this symptomatic data thus constitutes a consistent database for training AI models, particularly through deep learning, to identify major diseases. This has been demonstrated in the work of 24 on the automatic identification of wheat diseases in Europe using AI algorithms. Also, based on the training of approximately 8,438 images of symptoms acquired on PlantVillage, 25 have set up a system for the automatic detection of diseases in vineyards and mango trees in India. This will constitute a database of the symptoms of the main diseases affecting mango trees and, above all, a significant step forward in overcoming the lack of phytosanitary expertise and speeding up decision-making in the field by producers.Best model for automatic detection of major mango tree diseasesThe performance achieved by the V2-3 M, V3-3 M, and V4-3 M models demonstrates differentiated capabilities in detecting the symptoms of major mango tree diseases from annotated images with accuracies greater than 50%. Although the Kruskal–Wallis test did not reveal statistically significant differences among the trained models, descriptive performance metrics consistently favored the V2-3 M model. The absence of statistical significance is likely related to the limited number of independent training repetitions, which reduced the statistical power of the comparison. Future studies using larger datasets and a greater number of independent training runs are expected to provide a more robust statistical assessment.The V2-3 M model, with a mAP@50 of 50.3%, appears to be the most effective, confirming the importance of good parameterization and optimized architecture in automatic detection tasks. Its performance combines relatively balanced precision (55.1%), recall (49.1%), and F1-score (51.9%), indicating that it can be used as a starting point for a functional prototype of a mobile plant health diagnosis tool26. These metrics show a good compromise between the model’s ability to correctly locate symptoms and its ability to minimize both false positives and false negatives.On the other hand, the poor performance of the V3-3 M model (mAP@50 = 23.5%; recall = 26.7% ; F1-score = 36.3%) despite similar accuracy (56.8%) highlights limitations in the generalization of learning. This model seems to struggle to capture symptomatic diversity, particularly due to reduced sensitivity to morphological and chromatic variations in symptoms.The metrics of the V4-3 M model, which are intermediate (mAP@50 = 46.8%; F1-score = 48.4%) and reflect a certain robustness, albeit less marked than V2-3 M, show that improved performance depends on a compromise between architectural complexity, annotation quality, and training image diversity. This finding is consistent with the observations of27, who showed that training models on highly diversified image databases significantly improves the robustness of artificial intelligence tools applied to agriculture.Furthermore, the satisfactory performance of the V2-3 M model is in line with recent research into leaf disease detection using deep learning. For example, 28 achieved a mAP of 51% on tomato diseases using CNN architectures. 12 achieved 99% accuracy for a cashew anthracnose diagnosis model using tools such as Mobilenet and Inception. In terms of performance, the V2-3 M model is the most effective. Its ability to balance accuracy (55.1%) and recall (49.1%), combined with a mAP@50 of 50.3% and F1-score of 51.9%, makes it the most sui. for initial operational application. The V4-3 M model, although fairly close, remains slightly less robust, while V3-3 M should be ruled out due to its insufficient results. Thus, V2-3 M provides a solid basis for an automated diagnostic system, with prospects for improvement.These results confirm the feasibility of using artificial intelligence models for the automatic detection of major mango tree diseases, while highlighting the need to continue optimizing algorithms and enriching training databases29.The integration of such a tool into a mobile application could strengthen early detection capabilities and improve the responsiveness of phytosanitary interventions in Burkina Faso.Furthermore, the inference evaluation conducted on an independent dataset of 115 images from the Bobo-Dioulasso locality not used during training yielded a detection accuracy of 90.4% at a 70% confidence threshold. This result partially addresses the limitations of internal validation only, demonstrating that the V2-3 M model retains satisfactory detection capability when applied to new, unseen mango trees under comparable agroclimatic conditions. However, cross-regional validation across more diverse environmental conditions remains a perspective for future work to further consolidate the robustness and transferability of the model.The present study focused exclusively on the YOLOv11 architecture as a first step toward developing an automated diagnostic tool for mango tree diseases in Burkina Faso.This choice was deliberate. YOLOv11 was selected as a starting point given its recognized efficiency for real-time object detection on mobile devices, which aligns with the intended deployment context of a field-usable mobile application13,19. Systematic comparison with alternative architectures such as R-CNN, EfficientDet, SSD or ResNet-based classifiers remains a key perspective for future work and will allow a more rigorous assessment of the optimal approach for mango disease detection under West African field conditions.ConclusionAnthracnose, bacterial disease, and dieback are major phytosanitary constraints affecting mango trees. Unfortunately, producers have little knowledge of the epidemiological aspects of these diseases or the control techniques available. This study was conducted with the aim of designing an accurate automatic diagnostic model for the main diseases affecting mango trees using artificial intelligence. Training the collection of symptomatic images showed that the V2-3 M model performed better in terms of metrics for automatic disease detection.These relevant results open up new prospects for integrating AI tools into agroecological practices, particularly for phytosanitary monitoring and management in orchards. YOLOv11 demonstrated promising performance for mango disease detection in our dataset. Future work will focus on enriching the training dataset, comparing YOLOv11 with alternative deep learning architectures (R-CNN, EfficientDet, SSD, ResNet-based classifiers), and conducting cross-regional validation to further consolidate model robustness and transferability.It would be wise to integrate treatment recommendations into the model, deploy it on a mobile application, and promote it for the benefit of producers, all of which will contribute to effective disease management in mango orchards.Data availabilityData from this study will be made available to the journal as needed via the corresponding author’s e-mail address.Code availabilityNo custom code or novel computational pipeline was developed in this study. Image annotation and model training were performed using the Roboflow platform using the pretrained YOLOv11 framework. Therefore, no custom source code is associated with this study.ReferencesFAO. World Food and Agriculture - Statistical Yearbook (FAO (Rome, 2021).APEX & Fiche Sectorielle Mangue Du Burkina Faso. (2018). https://www.apexb.bf/catalogue-de-produits/mangue doi:https://www.apexb.bf/catalogue-de-produits/mangue?MARAH & Filière Mangue: Les Acteurs Font Le Bilan de La Campagne. (2023).Zida, I. et al. Effectiveness of Four Integrated Pest Management Approaches in the Control of Fruit Flies (Diptera: Tephritidae) in Mango Agro-Ecosystems in the South-Sudanian Zone of Burkina Faso. Adv. Ento- mology. 11, 124–142 (2023).Article Google Scholar Bougoum, H. et al. Caractérisation de Colletotrichum spp agent pathogène de l ’ anthracnose du manguier (Mangifera indica L.) au Burkina Faso. J. Appl. Biosci. 190, 19999–20020 (2023).Article Google Scholar Zombré, C. Diversite genetique et pathologique de xanthomonas citripv. mangiferaeindicae, bacterie responsable de la maladie des taches noires du manguier au burkina faso et en afrique de l’ouest » (Université Joseph Ki-Zerbo, Ouagadodugou, 2016).Dianda, Z. O. et al. Prévalence du dessèchement du manguier et evaluation de la fréquence des champignons associés à la maladie au Burkina Faso. J. Appl. Biosci. 126, 12686 (2018).Article Google Scholar Dembele, D. D. & Camara, B. Evaluation pré et post-récolte de l ’ incidence et de la sévérité de l ’ anthracnose du manguier au nord de la Côte d ’ Ivoire. Publ du CIRAD 1–2 (2019).Paul-alfred, K. K. Profitability and Competitiveness of the Mango Sector in Ivory Coast. Int. J. Econ. Financ Res. 10, 12–16 (2024).Article Google Scholar PIP COLEACP. Nouveaux ravageurs et maladies invasives. Bactériose du manguier provoquée par Xanthomonas citri pv. mangiferaeindicae. COLEACP PIP 1–14 p. doi: (2012). www.coleacp.org/pipLahrache, R. l’ impact de l ’ intelligence artificielle sur la prise de décision The Impact of Artificial Intelligence in Decision Making. 7, 660–678 (2024).Jayaprakash, V. & Rajagopal, M. K. cashew dataset generation using augmentation and ralsgan and a transfer learning based tinyml approach abstract: arXiv Prepr. arXiv2304.08766 (2023).Tirkey, D., Singh, K. K. & Tripathi, S. Performance analysis of AI-based solutions for crop disease identification, detection, and classification. Smart Agric. Technol. 5, 100238 (2023).Article Google Scholar Drabo, C., Sanou, J., Nikiema, Z., Dao, A. & Sawadogo, M. Diversité variétale des manguiers (Mangifera indica L.) et des pratiques culturales associées à la production au Burkina Faso Varietal diversity of mango trees (Mangifera indica L.) and agriculturals practices associated with production in Burkina Faso. 16, 787–797 (2022).ANAM. Etat du climat au Burkina Faso, Annee. https://meteoburkina.bf (2024) doi: (2024). https://meteoburkina.bfINSD. 5ième recensement général de la population et de l’habitat. (2022). www.insd.bf doi:www.insd.bfGagnevin, L. & Pruvost, O. Epidemiology and control of mango bacterial black spot. Plant. Dis. 85, 928–935 (2001).Article PubMed Google Scholar Ramos, L. J., Davenport, T. L., McMillan, R. T. & Lara, S. P. The resistance of mango (Mangifera indica) cultivars to tip dieback disease in Florida. Plant. Dis. 81, 509–514 (1997).Article PubMed Google Scholar Hussain, R. K. and M. YOLOV11: an overview of the key architectural enhancements. 1–9 (2024). (2024).Dembele, D. D., Og, L., Georges, D. & Amari, E. Pre and postharvest assessment of mango anthracnose incidence and severity in the north of Côte d ’ Ivoire Pre and postharvest assessment of mango anthracnose incidence and severity in the north of Côte d ’ Ivoire. Int. J. Biol. Chem. Sci. 14. https://doi.org/10.4314/ijbcs.v13i6.24 (2020).Dofuor, A. K. et al. Mango anthracnose disease: the current situation and direction for future research. Front. Microbiol. 14, 1–18 (2023).Article Google Scholar Kankam, F. & Adomako, J. Anthracnose Disease of Mango: Epidemiology, Impact and Management Options. (2022). https://doi.org/10.5772/intechopen.105934Dianda, O. Z. et al. Lasiodiplodia species associated with mango (Mangifera indica L.) decline in Burkina Faso and influence of climatic factors on the disease distribution. Physiol. Mol. Plant. Pathol. 126, 9 (2023).Article Google Scholar Picon, A. & Alvarez-Gila, A. C, M. S. & Amaia Ortiz-Barredo, Jone Echazarra, A. J. Réseaux neuronaux convolutifs profonds pour la classification des maladies des cultures à l’état sauvage à partir d’appareils de capture mobiles. ScienceDirect 161, 280–290 (2020).Rao, U. S., Swathi, R., Sanjana, V., Arpitha, L. & Chandrasekhar, K. Deep Learning Precision Farming: Grapes and Mango Leaf Disease Detection by Transfer Learning. Glob Transitions Proc. 2, 535–544p (2021).Article Google Scholar Elkadi, K. Expérimentation d’un modèle de détection précoce des maladies de la tomate par apprentissage profond. Rev. Marocaine Prot. des. Plantes. 14, 19–30 (2020).Google Scholar Ramcharan, A., Mccloskey, P., Baranowski, K. & Mbilinyi, N. A. Mobile-Based Deep Learning Model for Cassava Disease Diagnosis. Front. Plant. Sci. 10, 1–8 (2019).Article Google Scholar Boulent, J., Foucher, S. & Théau, J. Convolutional Neural Networks for the Automatic Identification of Plant Diseases. Front. Plant. Sci. 10, 15 (2019).Article Google Scholar Hossain, M. A., Sakib, S., Abdullah, H. M. & Arman, S. E. Deep learning for mango leaf disease identification: A vision transformer perspective. Heliyon 10, 36361 (2024).Article Google Scholar Download referencesAcknowledgementsThe research was conducted at the Regional Center of Excellence in Fruits and Vegetables (CRE/FL) based at the Institute of Environment and Agronomic Research (INERA) in Farako-bâ, Burkina Faso.FundingThis work benefited from the facilities of ACTS/RIMP/FONRID/INERA-01/2023 fund.Author informationAuthors and AffiliationsInstitute for the Environment and Agricultural Research, Central Horticulture Laboratory, National Center for Scientific and Technological ResearchFarako-Bâ Research Station, Bobo-Dioulasso, Burkina FasoZoéyandé Oumarou Dianda, Issaka Zida & Issa WonniHigher Institute for Sustainable Development, Yembila Abdoulaye Toguyeni University, Fada N’Gourma, Burkina FasoEmile Kaboré & Lankoandé BenjaminHigher School of Computer Science, Nazi Boni University, Bobo-Dioulasso, Burkina FasoZakaria Cheick Oumar Keita, Zakaria Kinda & Malo SadouanouanAuthorsZoéyandé Oumarou DiandaView author publicationsSearch author on:PubMed Google ScholarEmile KaboréView author publicationsSearch author on:PubMed Google ScholarZakaria Cheick Oumar KeitaView author publicationsSearch author on:PubMed Google ScholarIssaka ZidaView author publicationsSearch author on:PubMed Google ScholarZakaria KindaView author publicationsSearch author on:PubMed Google ScholarLankoandé BenjaminView author publicationsSearch author on:PubMed Google ScholarMalo SadouanouanView author publicationsSearch author on:PubMed Google ScholarIssa WonniView author publicationsSearch author on:PubMed Google ScholarContributions**DZO** : Conceptualization; Funding acquisition; Methodology; Resources; Roles/Writing - original draft, **KE: ** Manuscript’s writing, **DZO, KE, ZI, KZCO: ** Data collection, **KE** and **KZCO** : Initial processing, data treatement and analysis, **ZI, KZ, LB, MS, WI** : Critical revision and improvement of the scientific contentAll authors have read and approved the final version of the manuscript.Corresponding authorCorrespondence to Zoéyandé Oumarou Dianda.Ethics declarationsCompeting interestsThe authors declare no competing interests.Additional informationPublisher’s noteSpringer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Rights and permissionsOpen Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.Reprints and permissionsAbout this article