DriverGaze360: 1 Million Omnidirectional Gaze-Labeled Driver Frames

DriverGaze360 contains approximately 1 million gaze-labeled frames captured from 19 human drivers to support omnidirectional driver attention modeling. Developed by dfki-av and released around 2025, the data provides a 360-degree field of view for analyzing visual attention in driving scenarios. It includes object-level guidance to assist in mapping gaze to specific environmental elements.

Use Cases

• Training regression models to predict driver gaze coordinates within a 360-degree environment
• Developing object-level attention mechanisms for autonomous vehicle safety systems
• Analyzing temporal gaze shifts across sequential video frames to identify driver distraction

Strengths

• 1 million gaze-labeled frames
• 360-degree omnidirectional field of view
• Data collected from 19 unique human drivers

Limitations

• Small participant pool of 19 drivers may limit demographic and behavioral diversity
• Potential for label noise inherent in human gaze tracking

Provenance

Source

dfki-av

Collection Method

Sensor-based recording of human drivers using 360-degree cameras and gaze trackers

Freshness

Last updated March 2026.

Geography

United States