Využití generativních sítí pro analýzu řidiče ve vozidle

Abstract

This diploma thesis examines the use of generative networks for analyzing driver behavior in a vehicle. The main objective was to investigate the feasibility of using synthetic image data for a training set that could cover or supplement critical situations during driving that are difficult to capture. The first two theoretical chapters focus on introducing generative models, driver monitoring methods, and convolutional neural network architectures. The next two experimental chapters build upon the topics covered in the theoretical section. Both of these experimental sections conclude with a demonstration and description of the results.