基于深度学习混合模型迁移学习的图像分类

系统仿真学报 ›› 2016, Vol. 28 ›› Issue (1): 167-173.

基于深度学习混合模型迁移学习的图像分类

石祥滨^1,2,3, 房雪键³, 张德园¹, 郭忠强³

1.沈阳航空航天大学计算机学院,沈阳 110136;
2.沈阳航空航天大学辽宁通用航空重点实验室,沈阳 110136;
3.辽宁大学信息学院,沈阳 110036

收稿日期:2015-06-09 修回日期:2015-07-30 发布日期:2020-07-02
作者简介:石祥滨(1963-),男,辽宁,博士,教授,研究方向为虚拟现实、图像处理,网络游戏。
基金资助:
国家自然科学基金(61170185); 航空科学基金(2013ZC54011); 辽宁省博士启动基金(20121034); 辽宁省教育厅资助项目(L2014070)

Image Classification Based on Mixed Deep Learning Model Transfer Learning

Shi Xiangbin^1,2,3, Fang Xuejian³, Zhang Deyuan¹, Guo Zhongqiang³

1. Department of Computer, Shenyang Aerospace University, Shenyang 110136, China;
2. Liaoning General Aviation Key Laboratory, Shenyang Aerospace University, Shenyang 110136, China;
3. College of Information, Liaoning University, Shenyang 110036, China

Received:2015-06-09 Revised:2015-07-30 Published:2020-07-02

摘要/Abstract

摘要： 为提高深度模型迁移学习的特征识别力,提出一种基于受限玻尔兹曼机与卷积神经网络混合模型迁移学习的图像分类方法。该方法融合了2种模型特征的学习能力,提取图像的结构性高阶统计特征进行主题分类。该方法在迁移预训练的卷积神经网络模型到小目标集时,使用受限玻尔兹曼机代替卷积神经网络模型中的全连接层,在目标集上重新训练受限玻尔兹曼机层和Softmax层,并使用BP算法进行参数调整。加入的受限玻尔兹曼机层不仅全连接所有特征maps,还从最大对数似然的角度学习目标集特有的统计特征,消除了数据集间内容差异对迁移学习特征识别力的影响。在Pascal VOC2007和Caltech101数据集上的实验结果表明,该方法具有较高的分类准确率。

关键词: 图像分类, 卷积神经网络, 受限玻尔兹曼机, 迁移学习, Softmax

Abstract: In order to obtain high discrimination image representations in limited amount of datasets, the method based on mixed deep transfer learning model was proposed. When trained CNNs transferred to the target datasets, fully-connected layers were replaced by RBM layers. The method retrained the RBM layers and Softmax classifier, then fine-tuned the mixed model with backpropagation algorithm. The RBM layers not only fully connected whole feature maps, but also learned the target datasets' statistical features in the view of the biggest logarithmic likelihood, to eliminate the effects caused by the content differences between datasets. The experimental results show that the method has improved the accuracy of image classification, outperforming other methods on Pascal VOC2007 and Caltech101 datasets.

Key words: image classification, CNN(Convolutional Neural Networks), RBM(Restricted Boltzmann Machines), transfer learning, softmax

中图分类号:

TP391.9

石祥滨, 房雪键, 张德园, 郭忠强. 基于深度学习混合模型迁移学习的图像分类[J]. 系统仿真学报, 2016, 28(1): 167-173.

Shi Xiangbin, Fang Xuejian, Zhang Deyuan, Guo Zhongqiang. Image Classification Based on Mixed Deep Learning Model Transfer Learning[J]. Journal of System Simulation, 2016, 28(1): 167-173.

参考文献

[1] Fernández-Delgado M, Cernadas E, Barro S, et al.Do we Need Hundreds of Classifiers to Solve Real World Classification Problems?[J]. Journal of Machine Learning Research (S1532-4435), 2014, 15(1): 3133-3181.
[2] Joachims T.Making Large-scale Support Vector Machine Learning Practical [C]// Advances in kernel methods. USA: MIT Press, 1999: 169-184.
[3] Harada T, Ushiku Y, Yamashita Y, et al.Discriminative Spatial Pyramid[C]// Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on. USA: IEEE, 2011: 1617-1624.
[4] Sánchez J, Perronnin F, Mensink T, et al.Image Classification with the Fisher Vector: Theory and Practice[J]. International Journal of Computer Vision (S0920-5691), 2013, 105(3): 222-245.
[5] Zhang C, Liu J, Tian Q, et al.Image Classification by Non-negative Sparse Coding, Low-rank and Sparse Decomposition[C]// Computer Vision and Pattern Recognition (CVPR), 2011 IEEE Conference on. USA: IEEE, 2011: 1673-1680.
[6] Boureau Y L, Bach F, LeCun Y, et al. Learning mid-level features for recognition[C]//Computer Vision and Pattern Recognition (CVPR), 2010 IEEE Conference on. USA:IEEE, 2010: 2559-2566.
[7] Lecun Y, Kavukcuoglu K, Farabet C.Convolutional Networks and Applications in Vision[C]// Circuits and Systems (ISCAS), Proceedings of 2010 IEEE International Symposium on. USA: IEEE, 2010: 253-256.
[8] Bouvrie J.Notes on Convolutional Neural Networks [R]// MIT-CBCL Technical Reports. Germany: Springer International, 2006: 38-44.
[9] Fischer A, Igel C.Training Restricted Boltzmann Machines: An Introduction[J]. Pattern Recognition (S0031-3203), 2014, 47(1): 25-39.
[10] Tang Y, Salakhutdinov R, Hinton G.Robust Boltzmann Machines for Recognition and Denoising[C]// Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. USA: IEEE, 2012: 2264-2271.
[11] Le Q V, Ranzato M, Monga R, et al. Building High-level Features using Large Scale Unsupervised Learning[C]// Acoustics, Speech and Signal Processing (ICASSP), 2013 IEEE International Conference on. USA: IEEE, 2011: 8595-8598.
[12] He K, Zhang X, Ren S, et al. Spatial Pyramid Pooling in Deep Convolutional Networks for Visual Recognition[M]// Computer Vision - ECCV 2014. Germany: Springer International, 2014: 346-361.
[13] Krizhevsky A, Sutskever I, Hinton G E.Imagenet Classification with Deep Convolutional Neural Networks [C]// Advances in neural information processing systems. USA: The MIT Press, 2012: 1097-1105.
[14] Ciresan D, Meier U, Schmidhuber J.Multi-column deep neural networks for image classification[C]//Computer Vision and Pattern Recognition (CVPR), 2012 IEEE Conference on. USA:IEEE, 2012: 3642-3649.
[15] Ciresan D C, Meier U, Masci J, et al. Flexible, high performance convolutional neural networks for image classification[C]// IJCAI Proceedings-International Joint Conference on Artificial Intelligence. USA: Morgan Kaufmann , 2011, 22(1): 1237.
[16] Zeiler M D, Fergus R.Visualizing and understanding convolutional networks[M]//Computer Vision-ECCV 2014. Germany: Springer International Publishing, 2014: 818-833.
[17] Oquab M, Bottou L, Laptev I, et al. Learning and Transferring Mid-level Image Representations Using Convolutional Neural Networks[C]// 2014 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). USA: IEEE Computer Society, 2014: 1717-1724.
[18] Rumelhart D E, Hinton G E, Williams R J.Learning representations by back-propagating errors[J]. Nature (S0028-0836), 1986, 323(6088): 533-536.
[19] Hinton G E.Training Products of Experts by Minimizing Contrastive Divergence[J]. Neural Computation (S0899-7667), 2002, 14(8): 1771-1800.
[20] Song Z, Chen Q, Huang Z, et al. Contextualizing Object Detection and Classification[C]// Proceedings / CVPR, IEEE Computer Society Conference on Computer Vision and Pattern Recognition. IEEE Computer Society Conference on Computer Vision and Pattern Recognition. USA: 2011: 1585-1592.
[21] Yang J, Yu K, Gong Y, et al.Linear spatial pyramid matching using sparse coding for image classification[C]//Computer Vision and Pattern Recognition(CVPR), 2009 IEEE Conference on. USA:IEEE, 2009: 1794-1801.