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

This paper was presented in part at the CCF Chinese Conference on Computer Vision, Tianjin, 2017. This paper was recommended by the program committee.