A multi-class driver behavior dataset for real-time detection and road safety enhancement

Area of Science:

• Computer Science
• Artificial Intelligence
• Transportation Engineering

Background:

• Driver monitoring systems (DMS) are crucial for road safety.
• Existing datasets may lack diversity in real-world driving conditions and behaviors.
• Distracted driving remains a significant cause of traffic accidents.

Purpose of the Study:

• To introduce a novel, comprehensive dataset for training and evaluating AI-driven driver monitoring systems.
• To facilitate the development of intelligent transportation systems (ITS) for enhanced road safety.
• To support research in real-time driver behavior detection.

Main Methods:

• Collected 7286 high-resolution images of driver behaviors in diverse conditions (private vehicles, public buses) in Dhaka, Bangladesh.
• Categorized images into five classes: Safe Driving, Talking on the Phone, Texting, Turning, and Other Distracting Behaviors.
• Ensured dataset reflects natural variations in lighting, angles, and vehicle types for real-world applicability.

Main Results:

• A publicly available, annotated dataset of real-world driver behaviors has been created.
• The dataset captures natural variations essential for robust AI model training.
• It provides a valuable resource for advancing AI in driver monitoring.

Conclusions:

• The novel dataset is a significant contribution to the field of AI-driven driver monitoring systems.
• Availability of this dataset will accelerate the development of safer intelligent transportation systems.
• This resource is expected to aid in reducing accidents attributed to distracted driving.