Attention for Vision-Based Assistive and Automated Driving: A Review of Algorithms and Datasets

Kotseruba, Iuliia; Tsotsos, John K. · 2022 · OpenAlex-citations

DOI: 10.1109/tits.2022.3186613

archive: archived pipeline: cataloged verified

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

Summary

This review paper addresses the critical role of visual attention in enhancing road safety through vision-based assistive and automated driving systems. Motivated by the prevalence of driver inattention as a primary cause of accidents, the authors examine methods for modeling and detecting the spatio-temporal aspects of drivers' gaze—specifically where and when they look. The study focuses on two main application categories: Driver Monitoring Systems (DMS), which assess driver state to prevent accidents, and Advanced Driver Assistance Systems (ADAS) or autonomous vehicles, which utilize attention models to improve perception and decision-making. The review synthesizes research from 2010 to 2021, selecting 204 relevant papers from premier venues in intelligent transportation, robotics, and computer vision. The methodology involves a comprehensive literature search and categorization of algorithms and datasets. The authors provide a theoretical background on human visual attention, distinguishing between bottom-up (saliency-driven) and top-down (task-driven) mechanisms, and defining inattention types such as restricted, misprioritized, and diverted attention. They analyze machine learning approaches for in-vehicle gaze estimation, distraction detection, and drowsiness detection. For gaze estimation, the review evaluates feature extraction pipelines using facial landmarks, head pose, and eye coordinates, comparing traditional classifiers with deep learning models like Convolutional Neural Networks (CNNs). For distraction and drowsiness detection, the authors assess algorithms that utilize visual features, vehicle telemetry, and physiological signals, noting the shift from hand-crafted features to end-to-end deep learning architectures. Key findings indicate that while eye trackers offer high precision, camera-based systems are more practical for widespread deployment despite lower accuracy. CNN-based models generally outperform traditional methods in distinguishing between adjacent areas of interest (AOIs) within the vehicle, though challenges remain in differentiating subtle gaze shifts, such as those between the windshield and speedometer, and in handling occlusions like eyewear. Distraction detection algorithms frequently employ Support Vector Machines or recurrent neural networks, leveraging temporal data to identify non-driving-related tasks. However, the review highlights significant limitations in current research, including the lack of standardized evaluation metrics, the scarcity of diverse and high-quality public datasets, and the difficulty of validating simulator-based results against real-world on-road conditions. The significance of this work lies in its systematic organization of the state-of-the-art in attention-based driving technologies, providing a curated list of publicly available datasets and algorithms. The authors conclude that while progress has been made, broader deployment requires addressing data availability, improving the generalizability of models across different drivers and environments, and developing more robust evaluation frameworks. By outlining these challenges and potential solutions, the paper serves as a foundational resource for researchers aiming to develop safer, more reliable vision-based assistive and automated driving systems.

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.

Summary generated by qwen3.6-27b-prismaquant on 2026-06-26; verification: verified.

Topics

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

• gaze based attention detection
• distraction detection algorithms
• situational awareness
• attention allocation
• attention
• peripheral attention

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).

• Methodological Resource: tool software, dataset resource
• Theoretical Contribution: theory or model