Services on Demand
Journal
Article
Indicators
- Cited by SciELO
- Access statistics
Related links
- Cited by Google
- Similars in SciELO
- Similars in Google
Share
Revista U.D.C.A Actualidad & Divulgación Científica
Print version ISSN 0123-4226
Abstract
LOZADA-PORTILLA, William Alexander; SUAREZ-BARON, Marco Javier and AVENDANO-FERNANDEZ, Eduardo. Application of convolutional neural networks for detection of the late blight Phytophthora infestans in potato Solanum tuberosum. rev.udcaactual.divulg.cient. [online]. 2021, vol.24, n.2, e1917. Epub Dec 06, 2021. ISSN 0123-4226. https://doi.org/10.31910/rudca.v24.n2.2021.1917.
The presence of late blight in potato crops directly affects plant growth and tuber development; therefore, early detection of the disease is important. Currently, the application of convolutional neural networks is an opportunity oriented to the identification of patterns in precision agriculture, including the study of late blight in potato crops. This study describes a deep learning model capable of recognizing late blight in potato crops by means of leaf image classification. The PlantVillage augmented dataset was used in the application of this model for training. The proposed model has been evaluated from performance metrics such as precision, sensitivity, F1 score, and accuracy; to verify the effectiveness of the model in the identification and classification of late blight and compared in performance with architectures such as AlexNet, ZFNet, VGG16, and VGG19. The experimental results obtained with the selected data set showed that the proposed model achieves an accuracy of 90 % and an F1 score of 91 %. Therefore, it is concluded that the proposed model is a useful tool for farmers in the identification of late blight and scalable to mobile platforms due to the number of parameters that comprise it.
Keywords : Convolutional neural networks; Deep learning; Late blight; Precision agriculture.