Modelo de Redes Neuronales Convolucionales para detectar enfermedades en las hojas del cultivo de Quinua (Chenopodium quinoa) en el Centro Agronómico K’ayra, San Jeronimo, Cusco 2023

Abstract

In the world, crop diseases are the main cause of reduction in production quality. These diseases affect quinoa crops and a large amount of economic losses occur each year. It is essential to identify these diseases at an early stage to increase production. A visual inspection is the most common method to identify diseases, these errors are common through visual inspection. Time is a key factor in disease detection and requires experience. This study shows how image recognition can be used for disease detection. This work consisted of collecting a data set of images for leaf spot 1,120 images, for bacterial spot 850 images, for downy mildew 896 images and 1,090 healthy images for a total of 3,956 images of quinoa leaves from the K'ayra agronomic center in the Leticia sector, San Jeronimo, Cusco, Peru, of which 70% were considered for training, 20% for validation and 10% for testing. The proposed model worked correctly with an accuracy of 89.498%, which will allow quinoa farmers to detect diseases early, hopefully leading to an increase in quinoa production worldwide.

Plant phenotyping relevance

キヌア葉画像から病害状態を識別するCNNモデルの開発と精度評価が研究の中心であり、植物の病害表現型を直接推定している。

Code and data availability

The paper describes a self-collected quinoa leaf image dataset (3,956 images) and a CNN model, but provides no public deposit, repository, or availability statement for the dataset, code, or trained model. All listed URLs are cited references, not paper-specific assets.

No evidence-backed public reproduction asset is currently recorded.