CNN based Traffic Sign Classification using Adam Optimizer

An automatic detection and classification of traffic signs is an important task in Advanced Driver Assistance System (ADAS).Convolutional Neural Network (CNN) has surpassed the human performance and shown the great success in detection and classification of traffic signs. The paper proposes an approach based on the deep convolutional network for classifying traffic signs. The Belgium traffic sign dataset (BTSD) is used for evaluation and experiment results shows that the proposed method can achieve competitive results compared with state of the art approaches. Different activations and optimizers are used to evaluate the performance of proposed architecture and it is observed that Adam (Adaptive Moment Estimation) optimizer and softmax activation performs well.