Detect partial obstructions and dirty lenses before they impact safety. The enhanced Occlusion Detector identifies
significant frame blockages, while new distinct event types for "lens dirty" and "defocus" allow for precise maintenance
and root cause analysis.
Review evidence with greater clarity. Redesigned violation cards provide a clear distinction between incident types and feature the most relevant snapshots—such as a vehicle crossing entry/exit lines—to provide solid proof for justified responses.
The logic for detecting traffic violations has been improved to better account for complex vehicle maneuvers at difficult intersections and unexpected scenarios, such as traffic light outages.
Reduced Cross-Racial Effect In Demographic Classifications
Updates to the neural network model considerably improved the cross-racial misrecognition rate among ethnic demographic groups,
particularly between Caucasians and Asians.
This enhancement is reflected in a 2-point increase in the macro-F1 score.