SafeDriveMetrics - A Comprehensive Database for Visual, Cognitive, and Manual Distractions Using a Driving Simulator

The growing prevalence of driving distractions poses significant safety challenges. This paper introduces a database created using a driving simulator to mimic visual, cognitive, and manual distractions. The visual dataset includes videos of normal driving interspersed with distractions such as passenger interactions, billboard viewing, scenery observation, electronic display use, and rubbernecking. It features data from 30 subjects annotated with nine head poses, six eye gazes, and three mouth states. Cognitive distractions are simulated via math problem-solving and listening to songs, with high-speed facial videos capturing attentive versus distracted driving. The manual dataset includes images and videos categorized into three classes: both hands, one hand, and no hands on the steering wheel. This research aims to foster the development of algorithms to detect and classify all distractions. The dataset will serve as a benchmark to improve upon popular SFDDD and AUCDDD databases.