EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Image-Based Deep Learning Tool for Predicting Tomato Leaf Diseases

José Neftalí Torres Marín, Edward -J. Marín García, Andrés Felipe Calvo Salcedo

Tecnura · 31.03.2026 · 30 (87) · 3-24

DOI: 10.14483/22487638.23913 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Tecnura, online erschienen am 31.03.2026. Verfasst von José Neftalí Torres Marín und 2 weiteren Autor:innen. Die Arbeit nennt 24 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Objective

Late blight disease, caused by Phytophthora infestans, represents a critical threat to tomato cultivation, significantly affecting its quality and productivity. This study aimed to develop a technological tool for the early and precise detection of tomato disease in a controlled agricultural environment using machine learning (ML) techniques and digital image processing.. Methodology: A total of 3560 foliar images encompassing three severity levels of Phytophthora infestans were acquired. The Segment Anything Model (SAM) was implemented as a preprocessing step to isolate the biological tissue and eliminate chromatic background noise. Three classifiers were evaluated using a unified Monte Carlo cross-validation framework (10 iterations, 70/30 split): SVM and KNN, based on global chromatic descriptors (HSV/YIQ and PCA), compared with a CNN that extracts local spatial topologies.

Results

Within the evaluated dataset, the convolutional neural network (CNN) and the local distance-based model (KNN) achieved classification metrics exceeding 98%, significantly outperforming the global color-based baseline (SVM, 79%). By evaluating the localized spatial morphology of the necrotic lesions, the CNN architecture effectively mitigated misclassifications in extreme severity phenotype. Consequently, these high classification metrics must be interpreted as an upper performance bound, strictly constrained to the ideal and controlled conditions of the greenhouse setup.

Conclusions

The developed tool demonstrates high potential for automated disease diagnosis in tomatoes, contributing to phytosanitary management through emerging technologies. This study highlights ML’s role as a driver of innovation for more precise, efficient, and sustainable agriculture.

Zitieren

APA

Marín, J. N. T., García, E. J. M., & Salcedo, A. F. C. (2026). Image-Based Deep Learning Tool for Predicting Tomato Leaf Diseases. Tecnura, 30(87), 3-24. https://doi.org/10.14483/22487638.23913

Vancouver

Marín JNT, García EJM, Salcedo AFC. Image-Based Deep Learning Tool for Predicting Tomato Leaf Diseases. Tecnura. 2026;30(87):3-24. doi:10.14483/22487638.23913

BibTeX
@article{Marn2026,
  author = {Marín, José Neftalí Torres and García, Edward -J. Marín and Salcedo, Andrés Felipe Calvo},
  title = {Image-Based Deep Learning Tool for Predicting Tomato Leaf Diseases},
  journal = {Tecnura},
  year = {2026},
  volume = {30},
  number = {87},
  pages = {3--24},
  doi = {10.14483/22487638.23913}
}
Kostenlos starten →