Distractive driver behaviour detection with spiked neural networks

Loading...
Thumbnail Image

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis

Date

2022-11-01

Department

Major/Subject

Data Science

Mcode

SCI13115

Degree programme

Master's Programme in ICT Innovation

Language

en

Pages

33+5

Series

Abstract

Spiking neural networks are biologically plausible neural networks, capable of utilizing the concept of time throughout their simulations. They offer unique opportunities for the automotive industry because once they are deployed they require lower computational power and thus have decreased power consumption compared to other artificial neural networks. Theoretically, spiking neural network models can be just as powerful as the current state-of-the-art models, but in practice the technology is not matured enough to over-perform standard neural networks in terms of accuracy. The thesis examines a possible application of spiking neural networks to a real-life image classification problem, distractive drive behaviour detection. Detecting the distractive behaviours of drivers could prevent many accidents, thus saving lives. As the automotive industry moves towards even more automation to increase driving safety the need for low-power neural networks will significantly increase in the near future.

Description

Supervisor

Jung, Alexander

Thesis advisor

Tarsoly, Mate

Keywords

driver monitoring, distraction detection, machine learning, spiked neural networks

Other note

Citation