Um estudo sobre o TinyML a partir de uma aplicação mobile para detecção de doenças em folhas de manga

Abstract

In the global context, mango is one of the most important fruits in the world and, in Brazil, it also plays a very important role. However, diseases that affect mango trees can significantly harm the harvest of its fruits. In this sense, machine learning can be done to help the farmer find out if his mango tree suffers from any disease. From this, we have TinyML, which is a technology on the rise that allows the power of machine learning to be brought to hardware with limited resources and brings several advantages with its use, such as low cost, reduced latency , energy efficiency and safety. With this, this work proposed to bring a mobile application for detecting diseases in manga leaves and, in addition, to bring the impact of TinyML in comparison to the use of traditional machine learning models, normally hosted on some server in the cloud. It was noted, therefore, that the results obtained in this work corroborate the advantages of using this technology and that in situations where access to the internet is limited, this application can be of great value, because the use of the internet is not necessary.