Citas bibligráficas
Cipra, E., Rodríguez, C. (2023). Detección de Plagas y Enfermedades en los Cultivos de Maíz Utilizando Procesamiento de Imágenes con Redes Neuronales, en el Distrito de Cascas La Libertad, Año 2022 [Tesis, Universidad Privada Antenor Orrego]. https://hdl.handle.net/20.500.12759/14671
Cipra, E., Rodríguez, C. Detección de Plagas y Enfermedades en los Cultivos de Maíz Utilizando Procesamiento de Imágenes con Redes Neuronales, en el Distrito de Cascas La Libertad, Año 2022 [Tesis]. PE: Universidad Privada Antenor Orrego; 2023. https://hdl.handle.net/20.500.12759/14671
@misc{renati/371118,
title = "Detección de Plagas y Enfermedades en los Cultivos de Maíz Utilizando Procesamiento de Imágenes con Redes Neuronales, en el Distrito de Cascas La Libertad, Año 2022",
author = "Rodríguez Alva, Carlos Jhampiere",
publisher = "Universidad Privada Antenor Orrego",
year = "2023"
}
Farmers in the Cascas District do not have a procedure to help them prevent or detect in time the type of pest or disease that attacks their corn crops. The main objective of this thesis is to train and evaluate 4 models with neural networks such as ResNet50. EfficientNetVB0, MobileNetV2 and DenseNet201. For this purpose, a set of images were collected and classified into; images of Corn Leaf Blight (Blight), Common Rust (Common Rust), Armyworm (Armyworm), and plants with Healthy Leaves (Healthy). They were then grouped once again as training, validation and test images. For the development, augmentation techniques, filters and functions were applied. In order to improve the learning of neural network models. Then we proceed to measure how accurate the models are. Tests were carried out on corn leaves with worm, common rust, corn leaf blight and healthy leaves. The result was that the best model in terms of performance is DenseNet201. In the case of Pest Worm it has an accuracy of 0.97 which belongs to that category. For Common Rust with 0.95 accuracy. Leaf Blight at 0.98 accuracy. Healthy 0.73 accuracy. Emphasizing that these are unknown images for the models. To test the accuracy of the models we used the confusion matrix which measures how accurate their predictions are for each model. Giving MobileNetV2 as the model that has a better accuracy with respect to the other models with 99% certainty. As long as the training has a large number of images, filters, techniques and adequate functions. A model can give results that help to make better decisions because of the levels of accuracy that can be obtained.
Este ítem está sujeto a una licencia Creative Commons Licencia Creative Commons