Driver action monitoring based on convolutional neural network algorithms

Abstract

The paper substantiates the relevance of the task of monitoring the driver's actions, formulates the requirements for the implementation of such monitoring and proposes two variants of its implementation based on the use of a high-performance platform with built-in discrete graphics card and an ordinary cell phone. The obtained quantitative performance characteristics allow us to conclude about the practical possibility of solving the problem of recognition of the driver's actions in real time, including on low-performance platforms.