Aplicação mobile para detecção e diagnósticos de doenças em plantas usando TensorFow lite

Abstract

In order to improve productivity and help the farmer, a system was proposed in the present work, which will help the user in the detection of diseases and control, with their history and location,this is justified given the enormous diversity of diseases in plants and the difficulty in diagnosing them. and follow up on them. The system is capable of detecting previously cataloged diseases. From an image, which can be captured by the device’s camera, a processing is carried outwith the TensorFlow Lite tool, which will classify the image through a convolutional network model, which makes use of MobileNet, an architecture of neural network focused on mobile devices, having a smaller size and greater speed. By cataloging 24 classes, between healthy and diseased plants, the system achieved an average hit rate above 95% in certain cultures. The application is freely available under an open license on an open source platform.