terclim by ICS banner
IVES 9 IVES Conference Series 9 GiESCO 9 Implementation of a deep learning-based approach for detecting and localising automatically grapevine leaves with downy mildew symptoms

Implementation of a deep learning-based approach for detecting and localising automatically grapevine leaves with downy mildew symptoms


Context and purpose of the study – Grapevine downy mildew is a disease of foliage caused by Oomycete Plasmopara viticola an endoparasite that develops inside grapevine organs and can infect virtually every green organ. Downy mildew is one of the most destructive diseases in wine-growing regions, drastically reducing yield and fruit quality. Traditional manual disease detection relies on farm experts. Human field scouting has been widely used for monitoring the disease progress, however, is costly, laborious, subjective, and often imprecise. In this sense, computer vision technologies and artificial intelligence provide a suitable alternative to improve the current disease detection techniques. Therefore, this study aims to validate a deep learning-based approach for detecting and localising automatically leaves with downy mildew symptoms.

Material and methods – Fourteen commercial blocks (different grapevine cultivars) located in northern Spain were assessed to generate a comprehensive dataset for validating the deep learning algorithm. All analysed blocks presented downy mildew symptoms with different levels of intensity. RGB Images of the canopy were taken manually using a conventional camera. The images were acquired during different daily hours and under contrasting light conditions. YOLOv4 (You Only Look Once) was the deep learning algorithm analysed in this study. YOLOv4 model was trained using a heterogeneous dataset populated by RGB images obtained from the different vineyard blocks under different conditions to increase the robustness of the model. The RGB images were carefully labelled manually by an expert, selecting leaves with visible downy mildew symptoms. The labelled images were divided into two datasets, using 80% for training and validating and 20% for testing. The images used for training were divided into 1500 x 1500 pixel sub-images obtaining 15 sub-images per image and the sub-images were resized to 640×640 pixels. Data processing and deep learning modelling were performed with Python programming language and the Darknet neural network framework. The metrics used to evaluate the model were mean Average Precision (mAP), F1-score and Intersection over Union (IoU).

Results – The results of YOLOv4 for detecting leaves with downy mildew symptoms are promising. In the testing process applied on full images (2560×1728 pixels), the model presented a mAP of 67%, an F1-score of 0.69 and an IoU of 62%. When the number of real infected leaves (labelled by an expert) was compared with the predicted number of infected leaves the model reached a determination coefficient R2 of 0.93.  The accuracy of the method to determine the number of infected leaves was similar across the whole range of infections. This indicates that the model fits appropriately to all conditions tested. Also, the analysis of the localization indicates that leaves with apparent symptoms were detected correctly by the model.


Publication date: July 5, 2023

Issue: GiESCO 2023

Type: Poster


Carlos POBLETE-ECHEVERRÍA1,2, Inés HERNÁNDEZ1,2, Salvador GUTIÉRREZ3, Rubén ÍÑIGUEZ1,2, Ignacio BARRIO1,2 and Javier TARDAGUILA1,2*

1Televitis Research Group, University of La Rioja, 26006 Logroño, Spain
2Institute of Grapevine and Wine Sciences (University of La Rioja, Consejo Superior de Investigaciones Científicas, Gobierno de La Rioja), 26007 Logroño, Spain
3Department of Computer Science and Artificial Intelligence (DECSAI), Andalusian Research Institute in Data Science and Computational Intelligence (DaSCI), University of Granada (UGR), 18071 Granada, Spain

Contact the author*


artificial intelligence, deep learning, disease detection, precision viticulture, YOLOv4


GiESCO | GIESCO 2023 | IVES Conference Series


Related articles…

Tomatoes and Grapes: berry fruits with a (bright) biotech future?

Tomatoes and Grapes are berries that are genetically related and therefore at least partially their developmental pathways leading to a fleshy fruit should share some of the components. In a sense knowledge obtained from the model plant tomato could be useful for grape and conversely the more amenable tomato can be used to test some hypothesis that would be difficult to obtain in grape. Research in my lab and other labs have led to a better understanding of the molecular genetics mechanisms underlying fruit development and ripening in tomato and more specifically those related to metabolite accumulation that may lead to changes in fruit nutritional and flavor composition. This research has involved the use of genetic variability in natural population, but also biparental population and genetically engineered lines that are easy to develop in tomato tomato but not in grape. NGTs also can be easily implemented in tomato to not only speed up the gene-to-trait but also develop new tomato varieties.

Methodological advances in relating deep root activity to whole vine physiology

Full understanding of grapevine responses to variable soil resources requires
assessing the grapevine root system. Grapevine root systems are expansive and examining deep roots (i.e., >40 cm)
is particularly important in conditions where grapevines increase reliance on deep soil resources, such as drought
or plant competition. Traditional methods of assessing roots rely on morphological traits associated specific
functions (e.g., root color, diameter, length), while recent methodological advances allow for estimating root
function more directly (e.g., omics). Yet, the potential of applying refined methods remains underexplored for roots
at deep depths.

NACs intra-family hierarchical transcriptional regulatory network orchestrating grape berry ripening

Considering that global warming is changing berry ripening timing and progression, uncovering the molecular mechanisms and identifying key regulators governing berry ripening could provide important tools in maintaining high quality grapes and wine. NAC (NAM/ATAF/CUC) transcription factors represent an interesting family due to their key role in the developmental processes control, such as fruit-ripening-associated genes expression, and in the regulation of multiple stress responses. Between the 74 NAC family members, we selected 12 of them as putative regulators of berry ripening: NAC01, NAC03, NAC05, NAC11, NAC13, NAC17, NAC18, NAC26, NAC33, NAC37, NAC60 and NAC61.

Molecular characterization of a variegated grapevine mutant cv Bruce’s Sport

Variegation, a frequently observed trait in plants, is characterized by the occurrence of white or discoloured plant tissue. This phenomenon is attributed to genetic mosaicism or chimerism, potentially impacting the epidermal (L1) and subepidermal (L2) cell layers. In grapevine, variegation manifests as white or paler leaf, flower, or berry tissues, often leading to stunted growth and impeded development. Despite its prevalence, variegation in grapevines remains understudied.

Phenotypical impact of a floral somatic mutation in the cultivar Listán Prieto

The accession Criolla Chica Nº2 (CCN2) is catalogued as a floral mutation of cultivar Criolla Chica (synonym for cv. Listán Prieto). Contrary to what is observed in hermaphrodite-cultivated varieties like Criolla Chica, CCN2 exhibits a prevalence of masculinized flowers. Aiming to study the incidence and phenotypical implications of this mutation, CCN2 plants were deeply studied using Criolla Chica ‘Ballista’ (CCBA) as control plants. For each CCN2 plant, two inflorescences per shoot were sampled and segmented into proximal, mid and distal positions, relative to the pedicel. Flowers were observed through magnifying lens and classified according to OIV151 descriptor.