A Novel Approach for Detecting Traffic Signs using Deep Learning

Abstract

Nowadays, the demand for automatic driving assistance system is growing rapidly because it is reducing risk on drivers and helping to reduce the road accidents. In existing systems, many technological solutions are there but they are failing to produce promising accuracy. This paper has implemented a deep learning model for recognizing traffic signs that are present in the real world. The dataset that is utilized here is GTSRB which consists of 50,000 images of variable sizes. Due to modern technology and improvement in the automobile industry, numerous problems are encountering due to the huge number of vehicles. As a result, there is an increase in the number of accidents that happen due to the misreading of data by humans. To overcome the problem of misinterpretation, Traffic Sign Recognition is developed. Traffic Sign Recognition system capable of extracting traffic signs in real-time and can recognize the sign associated with the image. This model is being developed by using CNN. Our model producing 99.99% accuracy on training as well as validation data set. Traffic Sign Recognition is also a great contribution to the driver-less car technology that is being developed by Tesla. For a car to be driven without the help of a human, it should be able to detect traffic signs and act accordingly. The proposed model works effectively in different illuminating conditions and directions, where existing systems fail to produce promising results. This model helps to provide high accurate driver assisting system which can help to reduce accidents due traffic signal identification.