Traffic sign recognition based on deep convolutional neural network

Abstract

Traffic sign recognition (TSR) is an important component of automated driving systems. It is a rather challenging task to design a high-performance classifier for the TSR system. In this paper, we propose a new method for TSR system based on deep convolutional neural network. In order to enhance the expression of the network, a novel structure (dubbed block-layer below) which combines network-in-network and residual connection is designed. Our network has 10 layers with parameters (block-layer seen as a single layer): the first seven are alternate convolutional layers and block-layers, and the remaining three are fully-connected layers. We train our TSR network on the German traffic sign recognition benchmark (GTSRB) dataset. To reduce overfitting, we perform data augmentation on the training images and employ a regularization method named “dropout”. The activation function we employ in our network adopts scaled exponential linear units (SELUs), which can induce self-normalizing properties. To speed up the training, we use an efficient GPU to accelerate the convolutional operation. On the test dataset of GTSRB, we achieve the accuracy rate of 99.67%, exceeding the state-of-the-art results.

References

[1] NguwiY.Y., KouzaniA.Z.. Neural Computing and Applications, 2008, 17: 265

[2] MogelmoseA., TrivediM.M., MoeslundT.B.. IEEE Transactions on Intelligent Transportation Systems, 2012, 13: 1484

[3] LuX., WangY., ZhouX., ZhangZ., LingZ.. IEEE Transactions on Intelligent Transportation Systems, 2017, 18: 960

[4] LimK.H., SengK.P., AngL.M.. Intra Color-shape Classification for Traffic Sign Recognition, 2010, 642

[5] MadaniA., YusofR.. Neural Computing and Applications, 2017, 1

[6] LauM.M., LimK.H., GopalaiA.A.. Malaysia Traffic Sign Recognition with Convolutional Neural Network, 2015, 1006

[7] GlorotX., BordesA., BengioY.. Deep Sparse Rectifier Neural Networks, 2011, 315

[8] MaasA.L., HannunA.Y., NgA.Y.. Rectifier Nonlinearities Improve Neural Network Acoustic Models, 2013, 28: 6

[9] XuB., WangN., ChenT., LiM.. Empirical Evaluation of Rectified Activations in Convolutional Network, 2015,

[10] KlambauerG., UnterthinerT., MayrA., HochreiterS.. Self-Normalizing Neural Networks, 2017,

[11] LeCunY., BottouL., BengioY., HaffnerP.. Proceedings of the IEEE, 1998, 86: 2278

[12] KrizhevskyA., SutskeverI., HintonG.E.. Advances in Neural Information Processing Systems, 2012, 25: 1097

[13] SimonyanK., ZissermanA.. Very Deep Convolutional Networks for Large-scale Image Recognition, 2015,

[14] LinM., ChenQ., YanS.. Network in Network, 2014,

[15] SzegedyC., LiuW., JiaY., SermanetP., ReedS., AnguelovD., ErhanD., VanhouckeV., RabinovichA.. Going Deeper with Convolutions, 2015, 1

[16] HeK., ZhangX., RenS., SunJ.. Deep Residual Learning for Image Recognition, 2016, 770

[17] SzegedyC., IoffeS., VanhouckeV., AlemiA.A.. Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning, 2017, 4278

[18] StallkampJ., SchlipsingM., SalmenJ., IgelC.. Neural Networks, 2012, 32: 323

[19] AbadiM., AgarwalA., BarhamP., BrevdoE.. Tensorflow: Large-scale Machine Learning on Heterogeneous Distributed Systems, 2016,

[20] KingmaD., BaJ.. Adam: A Method for Stochastic Optimization, 2014,

[21] HintonG.E., SrivastavaN., KrizhevskyA., SutskeverI., SalakhutdinovR.R.. Improving Neural Networks by Preventing Co-adaptation of Feature Detectors, 2012,

[22] GlorotX., BengioY.. Understanding the Difficulty of Training Deep Feedforward Neural Networks, 2010, 249