Pedestrian Red-Light Violation Detection Based on Deep Learning Vision

With the increasing prevalence of cars and various modes of transportation in every household, pedestrians face growing potential dangers when crossing roads. Current traffic accident responsibilities often extend beyond drivers, as pedestrian traffic habits, such as jaywalking and distracted phone use, contribute to accidents. Against this backdrop, we have designed a real-time object detection system based on the YOLOv5 (You Only Look Once) algorithm, specifically tailored for detecting pedestrians violating red light signals. The system incorporates an alert mechanism to prompt pedestrians to return to a safe area promptly. YOLOv5 is renowned for its outstanding performance, flexible network architecture, and ease of deployment. This paper builds upon this foundation, aiming to reduce the occurrence of traffic accidents and ensure that pedestrians crossing roads receive timely safety alerts.