A Framework for Traffic Signs Recognition Using Convolutional Neural Network Over Conventional Models

In intelligent transportation systems, traffic sign recognition is crucial. Driver assistance and autonomous driving are two important functions of traffic sign detection. We present a Convolutional Neural Network model in this study that can identify and recognise traffic signals quickly and accurately. The proposed model is trained, tested and evaluated by using a dataset acquired from kaggle. The dataset is initially pre-processed for training the model, in the second stage the data is classified by using the proposed CNN model into different classes, the last stage of this model is to get the real time accuracy by testing the trained model. The CNN model was built using the Keras library. This model has shown significant results when tested against the dataset by achieving an accuracy of 98.87%.