Diseño de un prototipo web de recolección y validación de datos empleados para detectar enfermedades en las hojas de plantas medicinales a través de procesamiento de imágenes mediante redes neuronales convulsiónales caso de estudio: caléndula (Cercospora) y albahaca (Colletotrichum Gloeosporioides)

Abstract

In the field of computer vision, Convolutional Neural Networks (CNN) and the support of processing resources such Google Colab or computers with 24 or 32 GB ram, have made possible it to apply models to predict the classification of a given image. In the current document, previously trained models were used in the created web prototype for suggesting that an image contains a plant or not, and whether it shows a diseased or healthy plant. These automatic validations are mainly used to build subsequent training datasets, with similar topics or involving plants or plant leaves and their respective use and analysis. The web prototype was built with django, python and javascript. This work contains the application of these models within the web prototype, ensuring accessibility to the stored data, however it does not download the manual or user evaluation, which is also very important, since the models are sometimes over-adjusted.