Classifying Traffic Signs Using Convolutional Neural Networks Based on Deep Learning Models
Saira Akram, Sibghat Ullah Bazai, Shah Marjan · 2023
Deep learning (DL) is the sub-field of machine learning (ML), which is an emerging area of artificial intelligence (AI). DL techniques are being used in various applications of daily including object detection, data classification, segmentation, and many more. In the last decade, these techniques have been used to make predictions ranging from analysis of social networks, predicting and translation of sentences, and investments to life crucial problems like classifying images in healthcare and predicting traffic signs. The predictions done by these models are based on their learning algorithm and their past experience. Deep learning models are based on multi-layered artificial neural networks (ANNs) and the idea is perceived from the human brain. This chapter is written to impart all the basic concepts of ML needed to understand how a convolutional neural network (CNN) is used to classify the images of traffic signs. This chapter also highlights methods/algorithms used to train DL models and how the classification based on CNN can be more accurate and different from the traditional ML classification methods, along with most common pre-trained CNN models. The function of each layer in the architecture of CNN is evaluated in conjunction with the steps to classify the traffic signs and previously used methods are discussed thoroughly. Data sets consisting of multiple classes of traffic signs, including the German Traffic Sign Recognition Benchmark Dataset (GTSRBD), are compared comprehensively. Finally, the experimental setup providing basic guidelines in building a model to solve classification problems is also discussed to assist the researchers in the process of creating an efficient problem-solving model.