参数池化卷积神经网络图像分类方法

doi:10.3969/j.issn.0372-2112.2020.09.009

电子学报 ›› 2020, Vol. 48 ›› Issue (9): 1729-1734.DOI: 10.3969/j.issn.0372-2112.2020.09.009

参数池化卷积神经网络图像分类方法

江泽涛¹, 秦嘉奇², 张少钦³

1. 桂林电子科技大学广西图像图形与智能处理重点实验室, 广西桂林 541004;
2. 桂林电子科技大学信息科技学院, 广西桂林 541004;
3. 南昌航空大学, 江西南昌 330063

收稿日期:2018-12-27 修回日期:2020-04-19 出版日期:2020-09-25
- 通讯作者:
- 秦嘉奇
- 作者简介:
- 江泽涛 男,1961年3月出生,江西九江人.桂林电子科技大学教授,博士生导师,主要从事图像处理、计算机视觉、信息安全方面的研究.
- 基金资助:
- 国家自然科学基金 (No.61876049，No.61762066）; 广西科技计划 (No.AC16380108）; 广西图像图形智能处理重点实验项目 (No.GIIP201701，No.GIIP201801，No.GIIP201802，No.GIIP201803）; 广西研究生教育创新计划 (No.YCBZ2018052，No.2019YCXS043）

Parameterized Pooling Convolution Neural Network for Image Classification

JIANG Ze-tao¹, QIN Jia-qi², ZHANG Shao-qin³

1. Guangxi Key Laboratory of Image and Graphic Intelligent Processing, Guilin University of Electronic Technology, Guilin, Guangxi 541000, China;
2. Institute of Information Technology of Guet, Guilin, Guangxi 541000, China;
3. Nanchang Hangkong University, Nanchang, Jiangxi 330063, China

Received:2018-12-27 Revised:2020-04-19 Online:2020-09-25 Published:2020-09-25

摘要/Abstract

摘要： 传统的卷积神经网络使用池化层对信息进行降维操作，通常会造成信息损失，从而影响网络的表达能力.针对这一问题，使用参数池化层（Parameterized Pooling Layer）替代传统卷积神经网络中的池化层，提出参数池化卷积神经网络（Parameterized Pooling CNN，PPCNN）.参数池化层在仅仅增加了少量网络参数的情况下，最大可能的保留了卷积神经网络中希望被保留下来的特征；同时，由于增加了池化层前向传播的信息，从而影响了反向传播算法中权值的更新，网络收敛速度更快；实验结果表明，PPCNN模型与传统卷积神经网络模型以及部分改进模型相比，参数池化卷积神经网络模型是有效的.

关键词: 卷积神经网络, 图像分类, 池化方法, 参数优化

Abstract: Traditional convolutional neural network uses pooling layer to reduce the dimension of feature,which usually results in information loss,thus affecting the expression ability of the network.To solve this problem,the parameterized pooling layer is used to replace the pooling layer in the conventional convolutional neural network,and the parameterized pooling CNN (PPCNN) is proposed.In the case that only a few network parameters are added in the parameter pooling layer,it is possible to retain the desired features.At the same time,the forward propagation information of the pooling layer is added,which affects the update of weight in the backpropagation algorithm,and the network convergence speed is faster.Compared with the conventional convolutional neural network model and some improved models,experimental results show that the PPCNN model is effective.

Key words: convolutional neural network, image classification, pooling, parameter optimization

中图分类号:

TP391

江泽涛, 秦嘉奇, 张少钦. 参数池化卷积神经网络图像分类方法[J]. 电子学报, 2020, 48(9): 1729-1734.

JIANG Ze-tao, QIN Jia-qi, ZHANG Shao-qin. Parameterized Pooling Convolution Neural Network for Image Classification[J]. Acta Electronica Sinica, 2020, 48(9): 1729-1734.

参考文献

[1] Krizhevsky A,Sutskever I,Hinton G E.ImageNet classification with deep convolutional neural networks[A].International Conference on Neural Information Processing Systems[C].USA:Curran Associates Inc,2012.1097-1105.
[2] Simonyan,Karen,Zisserman,Andrew.Very deep convolutional networks for large-scale image recognition[J].Computer Vision and Pattern Recognition,2014,arXiv:1409.1556.
[3] Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[A].Proc of the IEEE Conf on Computer Vision and Pattern Recognition[C].Piscatawway,NJ:IEEE,2015.1-9.
[4] He K,Zhang X,Ren S,et al.Deep residual learning for imagerecognition[A].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition[C].USA:IEEE,2016.770-778.
[5] Howard A G,Zhu M,Chen B,et al.MobileNets:Efficient convolutional neural networks for mobile vision applications[A].IEEE Conference on Computer Vision and Pattern Recognition[C].Honolulu,Hawaii:IEEE,2017.1-14.
[6] Iandola F N,Han S,Moskewicz M W,et al.SqueezeNet:AlexNet-level accuracy with 50x fewer parameters and <0.5MB model size[A].International Conference on Learning Representations[C].Toulon,France:ICLR,2017.1-9.
[7] He K,Zhang X,Ren S,et al.Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2015,37(9):1904-1916.
[8] Yue K,Xu F,Yu J.Shallow and wide fractional max-pooling network for image classification[J].Neural Computing & Applications,2017(5):1-11.
[9] Lee C Y,Gallagher P,Tu Z.Generalizing pooling functions in CNNs:Mixed,gated,and tree[J].IEEE Transactions on Pattern Analysis & Machine Intelligence,2018,40(4):863-875.
[10] 罗畅,王洁,王鹏飞,等.卷积自编码器中粗粒度池化特征提取研究[J].电子学报,2017,45(10):2390-2401. LUO Chang,WANG Jie,WANG Peng-fei,et al.Coarse-grained pooled features learning in convolutional autoencoders[J].Acta Electronica Sinica,2017,45(10):2390-2401.(in Chinese)
[11] Memisevic R,Zach C,Hinton G E,et al.Gated softmax classification[A].Advances in Neural Information Processing Systems[C].Columbia,Canada:ACM,2010.1603-1611.
[12] Routray S,Ray A K,Mishra C,et al.Efficient hybrid image denoising scheme based on SVM classification[J].Optik,2018,157:503-511.
[13] Lee S J,Chen T,Yu L,et al.Image classification based on the boost convolutional neural network[J].IEEE Access,2018,99(6):12755-12768.
[14] 王强,李孝杰,陈俊.Supplement卷积神经网络的图像分类方法[J].计算机辅助设计与图形学学报,2018,(3):385-391. WANG Qiang,LI Xiao-jie,CHEN Jun.Image classification method based on supplement convolutional neural network[J].Journal of Computer-Aided Design & Computer Graphics,2018,(3):385-391.(in Chinese)
[15] Springenberg J T,Dosovitskiy A,Brox T,et al.Striving for Simplicity:The All Convolutional Net[J/OL].https://ui.adsabs.harvard.edu/abs/2014arXiv1412.6806S/abstract,2014/2019-04-06.

参数池化卷积神经网络图像分类方法

Parameterized Pooling Convolution Neural Network for Image Classification

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

编辑推荐

Metrics

[1]	吴靖, 叶晓晶, 黄峰, 陈丽琼, 王志锋, 刘文犀. 基于深度学习的单帧图像超分辨率重建综述[J]. 电子学报, 2022, 50(9): 2265-2294.
[2]	毛国君, 王者浩, 黄山, 王翔. 基于剪边策略的图残差卷积深层网络模型[J]. 电子学报, 2022, 50(9): 2205-2214.
[3]	袁海英, 曾智勇, 成君鹏. 面向灵活并行度的稀疏卷积神经网络加速器[J]. 电子学报, 2022, 50(8): 1811-1818.
[4]	王相海, 赵晓阳, 王鑫莹, 赵克云, 宋传鸣. 非抽取小波边缘学习深度残差网络的单幅图像超分辨率重建[J]. 电子学报, 2022, 50(7): 1753-1765.
[5]	张志文, 刘天歌, 聂鹏举. 基于实景数据增强和双路径融合网络的实时街景语义分割算法[J]. 电子学报, 2022, 50(7): 1609-1620.
[6]	丁毅, 沈薇, 李海生, 钟琼慧, 田明宇, 李洁. 面向CNN的区块链可信隐私服务计算模型[J]. 电子学报, 2022, 50(6): 1399-1409.
[7]	张波, 陆云杰, 秦东明, 邹国建. 一种卷积自编码深度学习的空气污染多站点联合预测模型[J]. 电子学报, 2022, 50(6): 1410-1427.
[8]	张云, 化青龙, 姜义成, 徐丹. 基于混合型复数域卷积神经网络的三维转动舰船目标识别[J]. 电子学报, 2022, 50(5): 1042-1049.
[9]	廖勇, 李玉杰. 一种轻量化低复杂度的FDD大规模MIMO系统CSI反馈方法[J]. 电子学报, 2022, 50(5): 1211-1217.
[10]	崔亚奇, 何友, 唐田田, 熊伟. 一种深度学习航迹关联方法[J]. 电子学报, 2022, 50(3): 759-763.
[11]	王雪松, 张翰林, 程玉虎. 基于自编码器和超图的半监督宽度学习系统[J]. 电子学报, 2022, 50(3): 533-539.
[12]	伍邦谷, 张苏林, 石红, 朱鹏飞, 王旗龙, 胡清华. 基于多分支结构的不确定性局部通道注意力机制[J]. 电子学报, 2022, 50(2): 374-382.
[13]	陈文俊, 杨春玲. 图像压缩感知的特征域优化及自注意力增强神经网络重构算法[J]. 电子学报, 2022, 50(11): 2629-2637.
[14]	肖斌, 陈嘉博, 毕秀丽, 张俊辉, 李伟生, 王国胤, 马旭. 基于一维卷积神经网络与循环神经网络串联的心音分析方法[J]. 电子学报, 2022, 50(10): 2425-2432.
[15]	周登文, 李文斌, 李金新, 黄志勇. 一种轻量级的多尺度通道注意图像超分辨率重建网络[J]. 电子学报, 2022, 50(10): 2336-2346.