Egocentric Officer-Civilian Interaction on Traffic Stops

Large-scale multimodal analysis of body-worn camera footage

AI-generated illustration of a traffic stop seen from an officer's body-worn camera

Dec 2024 – Present · Signal Analysis and Interpretation Laboratory (SAIL), USC

Body-worn cameras record traffic stops from the officer’s point of view. This creates a large, egocentric record of real-world interactions between officers and civilians. This project, run in partnership with law enforcement, builds video pipelines to analyze these interactions at scale.

Highlights

  • Designed a video embedding extraction pipeline and setting-selection process over 30K+ body-worn videos of traffic stops.
  • Analyzing 2K+ annotated videos of multimodal, collaborative officer-civilian activity from multiple views.