This dataset records granular vehicle behavior at signalized intersections across a major U.S. metropolitan area, captured by fixed camera systems. Each event documents discrete driving actions and maneuvers, real-time speed, direction of travel, and physical vehicle dimensions, with special focus on dilemma-zone decision events where drivers choose to stop or clear during phase changes. Pedestrian signal timing and intersection phase metadata are embedded to support multi-modal modeling. Spatial and temporal metadata enable mapping, proximity queries, and intersection-level trend analysis. The collection supports safety research, signal timing optimization, autonomous vehicle and ADAS training, freight behavior analysis, and edge-case modeling using real-world intersection footage and measurements.

Key attributes include:

• Event-level driving behavior and maneuver classification at intersections
• Dilemma-zone event identification and timing relative to signal phases
• Real-time speed, direction, and vehicle physical dimensions for class validation
• Embedded pedestrian signal timing and intersection phase metadata for multi-modal modeling
• Camera-captured imagery and spatial-temporal anchors for mapping, verification, and CV training

Traffic engineers, safety researchers, AV developers, and municipal planners can use this dataset to optimize signal timing, reduce intersection risk, and train models on real-world edge-case driver behavior at signalized crossings.