Developing a Neural Network Model to Classify Traffic Signs for Autonomous Vehicle System

Abstract

Recognition of traffic signs is relevant in many applications including self-driving car/driverless car, traffic mapping and traffic monitoring. Self-driving cars have the potential to revolutioniae urban mobility, with efficient, safe, easy and congestion-free transportability.As an AI implementation, this vehicle autonomy has many problems, such as the unfailing detection of traffic lights, signs, unclear lane lines, pedestrains, etc.., these problems can be solved due to the availability of graphical processing units(GPU) and cloud platform in the fields of deeplearning, computer vision.Deep learning models have recently demonstrated promientrepresentatin ability and achieved excellent success in traffic sign recognition.In this model we propose a model for effective traffic light detection and recognition using transfet learning, based on deep neural networks. The approach involves the use for learning transfer of the Tensors flow faster Region Convolutionary Neural Network Inception V2 model. The model was trained on a dataset containing various images of traffic signals in conjuction with Indian traffic signals which are classified into five class groups.