Aplicación móvil para la detección de 5 enfermedades en papas por medio de visión artificial en el sector Joyagshi.

Abstract

This research aimed to develop a mobile application to identify diseases in potato plants manifested in damage to their leaves through artificial vision and a neural network. The mobile application requirements were developed using Python and Android Studio software. This application consists of three buttons: detect, manual, and information. The tests carried out in the training of the neural network indicate that in epochs of 46, the algorithm stops training with a loss of 0.3757 for training and performance of 0.8564; for validation, there is a loss of 0.6180 and performance of 0.7560; these data indicate that the training of the network is appropriate. From the tests carried out, by comparing images with MATLAB software, it was verified that for the black surf disease, there is a 97.48% similarity in height and a 99.91% similar in width; for the virus disease, there is a 97.48% similarity in height and a 99.91% similarity in width, the early blight disease 71.48% similarity in height and 99.88% similarity in width, for the rust disease there is a 93.1% similarity in height and 99.83% similarity in width, for late blight disease there is a 79.49% similarity in height and 99.51% similarity in width. The tests carried out on high, medium, and low-range devices showed that the high-range device has higher performance. The tests concluded that the mobile device application has the correct functionality, and the cost is low compared to the Plantix application. It is recommended to continue researching other diseases in potato cultivation, monitoring them, updating information, and releasing new technologies.