Harnessing CNN Architecture for Accurate Traffic Sign Classification: Findings from the GTSRB Dataset

Traffic signs are vital components of road infrastructure, serving as visual cues that communicate essential information to drivers, pedestrians, and cyclists. They play a crucial role in maintaining road safety by guiding traffic, indicating hazards, and regulating speed limits and right-of-way. Additionally, traffic signs help to establish order and predictability on the roads, reducing the likelihood of accidents and promoting smooth traffic flow. The importance of traffic sign classification lies in its ability to enhance the accuracy and efficiency of sign recognition systems, contributing to safer driving environments and the development of intelligent transportation systems. In this research we have taken the GTSRB - German Traffic Sign Recognition Benchmark dataset and included 516 more images in the dataset. After some preprocessing steps, we split the dataset into train and test data. Finally, we have trained the data with GRU, VGG-16, VGG-19, and scratch CNN architecture. Accuracies found are 91.91%, 87.19%, 86.53%, and 99.62%.