Analysis of ESP32 SoC for Feed-Forward Neural Network Applications

In the last few years, microcontrollers used for IoT devices became more and more powerful, and many authors have started to use them in machine learning systems. Most of the authors used them just for data collecting for ML algorithms in the clouds, but some of them implemented ML algorithms on the microcontrollers. The goal of this paper is to analyze the speed of one popular SoC (Espressif ESP32) running a neural network application. Neural networks with one and two hidden layers are used with the different number of neurons and the different number of inputs (9, 36, 144 and 576). This SoC has been analysed because some companies started to produce them with UXGA camera implemented and it can be used to distribute computing load from central ML servers to them. They can be so-called “edge” devices and in this paper speed of ESP32 SoC in feed-forward mode has been analysed.