Abstract

Accurately predicting the intentions of drivers during their maneuvers and assessing their level of attentiveness hold paramount importance in comprehending driver behavior and enhancing decision-making processes. Given the intricate nature of driving environments, drivers are contended with a plethora of distractions that divert their focus away from the road. Thus, it is imperative to fathom the extent of a driver's attentiveness and its consequential impact on their cognitive decision-making abilities in the context of making driving maneuvers. Despite its pivotal significance, research in the following fundamental questions remains conspicuously absent: What garners the primary visual focus of drivers whilst operating a vehicle? Why do drivers occasionally veer into other vehicles or obstacles? The current study therefore alleviates the gaps in previous studies by providing answers to these questions. The overarching goal of this study is to employ a zero-shot learning technique to understand the complex patterns and relationships between driver’s attention and their cognitive decision-making. To achieve this goal, the study leveraged on the synergistic fusion of both image features from the driving scene and eye gaze information from drivers. The data used from this study was gotten from DR(eye)VE dataset. We employ the Segment Anything Model (SAM) for our zero-shot learning to segment the attention of the drivers base on their gaze information. The result from our study indicates that drivers pay more attention to road when driving on a highway as compared to city drive. This insight provides valuable information about drivers' attention allocation in different driving contexts and can aid in developing tailored interventions to improve road safety and enhance driver awareness.