A Review of Driver Gaze Estimation and Application in Gaze Behavior Understanding

Sharma, Pavan Kumar; Chakraborty, Pranamesh · 2023 · arXiv

archive: archived pipeline: cataloged verified

Get this paper ↗ (full text — opens at the source; we link to it, we don't host it)

Summary

This review paper addresses the critical need for comprehensive understanding of driver gaze estimation and its applications in enhancing road safety. Motivated by the high global mortality rate from road crashes, largely attributed to driver distraction, drowsiness, and inattentiveness, the authors aim to synthesize existing literature on gaze fundamentals, estimation methods, and behavioral applications. The study fills gaps in previous reviews by providing an in-depth analysis of metrics, feature extraction, and algorithms specific to driving contexts, as well as exploring gaze behavior in complex scenarios like intersections and lane changes. The authors conducted a systematic literature review following PRISMA 2020 guidelines, searching Google Scholar and Scopus for papers published up to June 2023. After screening 1,100 initial records and applying strict inclusion criteria focused on passenger car studies, 154 relevant papers were selected for analysis. The review categorizes driver gaze representation into three methods: zone-based classification (discrete areas like windshield or mirrors), gaze direction/vector (3D spatial coordinates), and gaze object identification (specific traffic entities). It contrasts two primary estimation setups: head-mounted eye trackers, which offer high accuracy and fine-grained metrics like fixation and saccades but are intrusive and require calibration; and remote camera systems, which are non-intrusive and comfortable but suffer from lower accuracy and sensitivity to lighting and head movement. The paper also details data collection methodologies, distinguishing between stationary vehicle setups, which allow controlled labeling but lack generalizability, and moving vehicle setups, which provide realistic data but pose labeling challenges. Key findings highlight the evolution of gaze estimation from traditional machine learning to deep learning techniques, utilizing datasets collected via RGB, infrared, and RGB-D cameras. The review identifies specific temporal metrics crucial for behavior understanding, such as dwell time, glance duration, and entropy rates, which quantify driver attentiveness and mental workload. It emphasizes that while head-mounted systems capture precise eye movements like pupil dilation and saccades, remote systems primarily rely on coarser glance transitions between predefined gaze zones. The analysis reveals that driver gaze data is increasingly applied to build advanced driver assistance systems (ADAS), predict maneuvers, and detect inattention or distraction. However, significant challenges remain, including the lack of generalization from stationary to moving vehicle data, illumination vulnerabilities, and the intrusiveness of high-accuracy wearable devices. The significance of this work lies in its comprehensive synthesis of driver gaze technologies, providing a clear roadmap for researchers and engineers. By clarifying the trade-offs between accuracy and usability in different estimation setups and standardizing terminologies, the paper supports the development of more robust driver monitoring systems. It underscores the importance of integrating gaze behavior understanding into safer road infrastructure design and autonomous driving technologies, while identifying future research directions to overcome current limitations in data collection and algorithmic robustness.

Key finding

Deep learning gaze estimation approaches outperform traditional machine learning under varied lighting and large head movements, and driver gaze is a tractable input for inattention, distraction, and maneuver-prediction systems, though dataset and generalization gaps remain.

Methodology

review

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed. Discovered via discover_arxiv_cs.HC on 2026-05-04 (5 acquisition events logged).

Summary generated by qwen3.6-27b-prismaquant on 2026-06-10; verification: verified_with_issues.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• gaze based attention detection
• eye movements scanning
• attention allocation
• looked but failed to see
• visual
• distraction detection algorithms

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: behavioral performance data
• Methodological Resource: tool software, measurement protocol