Bhaiya Vaibhaw Kumar supervised by Dr. Kavita Vemuri received his Master of Science – Dual Degree in Electronics and Communication Engineering (ECD). Here’s a summary of his research work on Gaze Behaviour of Indian Two-wheeler Drivers – an Eye-Tracking Study

Understanding driver gaze behavior is crucial for improving road safety, particularly in dynamic and heterogeneous traffic conditions. While extensive research exists on four-wheeler drivers in structured environments, limited work addresses two-wheeler gaze behavior in unstructured, high-density traffic such as that on Indian roads. This thesis presents a detailed study of the visual attention patterns of Indian two-wheeler riders and introduces the myEye2Wheeler dataset, a large-scale real-world eye-tracking dataset collected during urban navigation. Participants were categorized as novice or experienced riders, enabling comparative analysis of how expertise influences gaze fixation and attention allocation. A core objective of this thesis is to study rider behavior through eye movement data. Unlike drone or ego-centric cameras that capture only spatial or scene-level context, eye tracking reveals what the rider visually prioritizes in real time. This provides access to the underlying cognitive processes of attention, prediction, and perception. The dataset comprises over 261,000 frames of gaze-tracked video under diverse traffic conditions, including interactions with pedestrians, sudden lane changes, and road hazards such as potholes. Results show that experienced riders use a more predictive gaze strategy, scanning potential hazard zones, whereas novices show reactive patterns, fixating on immediate threats. This reflects a distinction between top-down and bottom-up attention: experienced riders rely on contextual cues and prior knowledge to anticipate events, while novices respond more directly to stimuli. These findings align with cognitive theories of expertise and situational awareness. The thesis further examines how gaze behavior is influenced by environmental factors like traffic density and vehicle composition. Statistical insights into fixation patterns across object categories – vehicles, pedestrians, and on/off road support the design of more intuitive Advanced Driver Assistance Systems (ADAS) for two-wheelers. These analyses reveal how attention shifts in complex traffic, informing the development of context-aware ADAS alerts and realistic training programs. By offering a first-of-its-kind dataset and analysis for two-wheeler drivers, this thesis addresses a major gap in attention modeling. The insights have implications for safety policy, rider training, and ADAS development, and lay the groundwork for advances in human-centric traffic systems and computational gaze analysis. 

July 2026