Bilateral Filtering NIN Network for Image Classification

A novel deep architecture bilateral filter NIN for classification tasks is proposed in the paper, in which the input image pixels using the bilateral filter and a multi-path convolution neural network are reconstructed. This network has two input paths, one is the original image and the other is the reconstructed image which independent on and complement each other. Therefore, the loss of foreground object texture and shape information can be reduced during the process of feature extraction from the complex background images. Then, the softmax classifier is employed to classify the extracted features. Experiments are demonstrated on CAFIR-100 dataset, in which some object’s feature gradually disappear after pass through a series of convolution layers and average pooling layers. The results show that, Compared with NIN(network in net- work), the classification accuracy rate increased 0.6% on CIFAR-10 database, accuracy rate increased 0.27% on cifar-100 database.

[1]  Rob Fergus,et al.  Visualizing and Understanding Convolutional Networks , 2013, ECCV.

[2]  Li Fei-Fei,et al.  ImageNet: A large-scale hierarchical image database , 2009, CVPR.

[3]  Qiang Chen,et al.  Network In Network , 2013, ICLR.

[4]  Pascal Vincent,et al.  Representation Learning: A Review and New Perspectives , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[5]  Ian J. Goodfellow.  Piecewise Linear Multilayer Perceptrons and Dropout , 2013, ArXiv.

[6]  Nitish Srivastava,et al.  Improving neural networks by preventing co-adaptation of feature detectors , 2012, ArXiv.

[7]  Dumitru Erhan,et al.  Going deeper with convolutions , 2014, 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[9]  Rob Fergus,et al.  Stochastic Pooling for Regularization of Deep Convolutional Neural Networks , 2013, ICLR.

[10]  Quoc V. Le,et al.  ICA with Reconstruction Cost for Efficient Overcomplete Feature Learning , 2011, NIPS.

[11]  Geoffrey E. Hinton,et al.  Rectified Linear Units Improve Restricted Boltzmann Machines , 2010, ICML.

[12]  Yoshua Bengio,et al.  Maxout Networks , 2013, ICML.

[13]  Andrew Y. Ng,et al.  Reading Digits in Natural Images with Unsupervised Feature Learning , 2011 .

[14]  Frédo Durand,et al.  A Fast Approximation of the Bilateral Filter Using a Signal Processing Approach , 2006, ECCV.

[15]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[16]  Yoshua Bengio,et al.  Gradient-based learning applied to document recognition , 1998, Proc. IEEE.