In-field disease symptom detection and localisation using explainable deep learning: Use case for downy mildew in grapevine
Computers and Electronics in Agriculture · 2024-11 · 226 · 109478
DOI: 10.1016/j.compag.2024.109478 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computers and Electronics in Agriculture, 2024-11. Verfasst von Inés Hernández und 4 weiteren Autor:innen. Die Arbeit wurde bisher 33-mal zitiert (Crossref) und nennt 49 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC-ND 4.0). Gefördert von: University of La Rioja, European Research Council, Horizon 2020.
Abstract (Originalsprache)
Diseases and pests in agriculture represent a major loss in crop yield and quality. One example is downy mildew (Plasmopara viticola), a key disease of grapevine. Traditional disease detection is usually performed manually by experts in the field, resulting in a subjective and time-consuming assessment, so a technological solution based on artificial intelligence, would provide higher levels of reproducibility and sampling. The aim of this work was to develop a method for automated detection and localisation of plant disease symptoms under field conditions. Images of the grapevine canopy were taken in 14 commercial vineyard plots, with different daylight conditions. The images were divided using a sliding window and these sub-images were classified into areas with and without downy mildew. Transfer learning, fine-tuning and data augmentation were used to automate the classification, comparing convolutional neural networks (CNNs) and vision transformers (ViT). Finally, areas with downy mildew symptoms were located in the canopy images by combining the trained model with the sliding window method. Model predictions were interpreted using Explainable Artificial Intelligence (XAI) methods. The EfficientNetX2S model achieved an accuracy of 91% and an F1-score of 0.92 when classifying image areas and an Intersection over Union (IoU) of 0.83 when locating symptomatic areas. This method showed promising results, enabling automatic and explainable detection and localisation of plant diseases in complex conditions, with the potential to manage them through a mobile platform and reduce their spread.
Förderung
- University of La Rioja
- European Research Council
- Horizon 2020
Zitieren
APA
Hernández, I., Gutiérrez, S., Barrio, I., Íñiguez, R., & Tardaguila, J. (2024). In-field disease symptom detection and localisation using explainable deep learning: Use case for downy mildew in grapevine. Computers and Electronics in Agriculture, 226, 109478. https://doi.org/10.1016/j.compag.2024.109478
Vancouver
Hernández I, Gutiérrez S, Barrio I, Íñiguez R, Tardaguila J. In-field disease symptom detection and localisation using explainable deep learning: Use case for downy mildew in grapevine. Computers and Electronics in Agriculture. 2024;226:109478. doi:10.1016/j.compag.2024.109478
BibTeX
@article{Hernndez2024,
author = {Hernández, Inés and Gutiérrez, Salvador and Barrio, Ignacio and Íñiguez, Rubén and Tardaguila, Javier},
title = {In-field disease symptom detection and localisation using explainable deep learning: Use case for downy mildew in grapevine},
journal = {Computers and Electronics in Agriculture},
year = {2024},
volume = {226},
pages = {109478},
doi = {10.1016/j.compag.2024.109478}
}
EUVIMED