Real-Time Two-Stage Detection of Indian Traffic Signboards Using YOLO11 on Jetson Orin Nano

This paper presents a distinct and efficient method for traffic signboard recognition, addressing the critical need for enhancing road safety in India. This paper aims to develop a system that recognizes and alerts drivers about traffic signboards in real-time, thereby reducing accidents caused by drivers’ inattentiveness to road signs. The methodology involves a two-stage detection and classification process utilizing the YOLO11(You Only Look Once) image processing model, implemented on the Jetson Orin Nano developer kit. Stage 1 focuses on detecting the presence of signboards, achieving a precision of 75.7% and mAP50 of 65.5%, while Stage 2 classifies the detected signs with an impressive precision of 80% and mAP50 of 90.3%. We have a near realtime working for the traffic signboard detection and classification with inference time around 10ms. We have achieved a comparable result with a high mAP50 score. The solution integrates YOLO11’s powerful object detection capabilities with the robust hardware performance of Jetson Orin Nano, offering a practical deployment strategy for modern vehicles.