No Pedestrian Left Behind: Real-Time Detection and Tracking of Vulnerable Road Users for Adaptive Traffic Signal Control

There’s this cool new traffic signal system called No Pedestrian Left Behind (NPLB) that uses real-time object detection and multi-object tracking to help protect vulnerable road users at crosswalks. It automatically extends the signal time when needed. In tests with the BGVP dataset, YOLOv12 achieved the best mean Average Precision at 50% (mAP@0.5) with a score of 0.756. NPLB merges YOLOv12, ByteTrack tracking, and a smart controller that extends pedestrian phases if the timer gets low. In 10,000 Monte Carlo simulations, it improved VRU safety by 71.4%, cutting down stranding rates from 9.10% to 2.60%, while needing only a few signal extensions. It really addresses the problem of fixed-timing signals that can leave vulnerable folks stranded.