Aprendizado de máquina em microcontroladores utilizando TinyML: exploração, otimização e portabilidade

Abstract

The TinyML technique refers to the set of approaches that enable the implementation of machine learning algorithms on devices with restricted computational resources and memory capacity, such as embedded systems. This work addressed two ways to implement such techniques as optimizations and model compression, exploring different technologies. Additionally, specific details related to this approach to TinyML in the development process were presented, with an emphasis on portability and scalability. The evaluation of the proposed solution will allow analyzing the impact and effectiveness of using TinyML in implementing machine learning systems on microcontrollers with limited resources.