Traffic Signal Recognition Using the Neuro-T Deep Learning Platform: A Case Study on Real-time Object Detection in Urban Environments

Abstract

This study investigates the effectiveness of the Neuro-T deep learning platform in detecting and classifying traffic signals from real-world urban environments. The accurate detection of traffic lights is crucial in autonomous driving systems and intelligent transportation networks. A dataset comprising 500 real-world traffic light images was annotated and trained using Neuro-T’s GUI-based platform, employing a CNN-based architecture. The model achieved a classification accuracy of 94.8% and operated in real-time at 43 frames per second (FPS). The results demonstrate Neuro-T's potential as a lightweight and highly usable tool for rapid deployment of computer vision systems in urban mobility contexts.