Traffic Sign Recognition Using Deep Learning Module

According to an Australian Bureau of Statistics survey, approximately 26 million vehicles will be sold in 2022. Due to the rapid increase in usage rate, the traffic on the roads and accident rate also increase. To reduce this, vehicles have been equipped with an automation system that will automatically recognize the traffic signal. Traffic sign recognition is a challenging task in the field of computer vision. It plays a significant role in the development of intelligent transportation systems. Most of the researchers have done their research in traffic sign recognition using deep learning techniques. The network like Alex Net, VGG16, and LeNET-5 has been proposed. For larger datasets, LeNET-5 networks provide better results, and for smaller datasets, the network has an inconsistency. In this proposed work, the size of the datasets is small. Hence, the network has an inconsistency. To make the network is a consistent more number of features need to be provided. For extracting more features from the image, the kernel size has to be reduced. Reducing the kernel size will lead to overfitting. In order to avoid overfitting, dropout functions are added to the LeNET-5 architecture. The proposed modified LeNET-5 model uses a kernel with a 3 × 3 size, and dropout functions are added. The proposed network model has been validated by using German traffic sign datasets. The proposed architecture provides improved performance over the existing model.