Traffic Sign Detection Using Deep Learning Approach

Abstract

Traffic sign detection stands as a vital operational aspect of autonomous driving systems while forming an essential part of intelligent transportation technology. Research investigations take a deep learning method to detect traffic signs in images. The collected traffic sign image dataset from Kaggle underwent a precursing process including resizing as well as contrast enhancement and gamma correction and data augmentation. The researcher established three sets for data processing: training with 3,200 images and testing with 400 images alongside validation with 400 images.Analysis of traffic sign recognition capabilities was conducted by evaluating multiple deep learning models namely VGG19 as well as ResNet together with Xception and DenseNet in addition to AlexNet. Xception demonstrated the best accuracy performance at 99% among the evaluated models and AlexNet placed second with 98%, DenseNet achieved 96% accuracy while VGG19 obtained 95% accuracy and ResNet reached a lower mark of 70%. This study establishes modern convolutional neural networks (CNNs) perform traffic sign classification tasks efficiently through which Xception delivered the research's best classification outcomes.Autonomous vehicle navigation and road safety improvements receive significant benefits thanks to deep learning's ability to identify traffic signs precisely during real-time operations. Additional research will concentrate on dataset enlargement along with performance optimization and implementation of the system for practical use