[1]汤 凯,何 庆,赵 群,等.基于改进的深度残差网络的图像识别[J].南京师范大学学报(自然科学版),2019,42(03):115-121.[doi:10.3969/j.issn.1001-4616.2019.03.015]
　Tang Kai,He Qing,Zhao Qun,et al.Image Recognition Based on Improved Deep Neural Network[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):115-121.[doi:10.3969/j.issn.1001-4616.2019.03.015]
点击复制

基于改进的深度残差网络的图像识别()
分享到：

《南京师范大学学报》（自然科学版）[ISSN:1001-4616/CN:32-1239/N]

卷:
第42卷
期数:
2019年03期
页码:
115-121
栏目:
·全国机器学习会议论文专栏·
出版日期:
2019-09-30

文章信息/Info

Title:
Image Recognition Based on Improved Deep Neural Network
文章编号:
1001-4616(2019)03-0115-07
作者:
汤 凯; 何 庆; 赵 群; 王 旭
贵州大学大数据与信息工程学院,贵州省公共大数据重点实验室,贵州 贵阳 550025
Author(s):
Tang Kai; He Qing; Zhao Qun; Wang Xu
College of Big Data and Information Engineering,Guizhou University,Guizhou Provincial Key Laboratory of Public Big Data,Guiyang 550025,China
关键词:
图像分类; 深度学习; 深度残差网络; 空间变换网络
Keywords:
image classificaiton; deep learning; deep residual network; spatial transformation network
分类号:
TP183
DOI:
10.3969/j.issn.1001-4616.2019.03.015
文献标志码:
A
摘要:
随着大数据时代的发展,深度学习也渐渐变得更加实用,引领人工智能时代的发展. 卷积神经网络在图像领域中发挥着非常重要的作用,是深度学习模型中重要组成部分之一. 图像识别的关键攻破点在于如何提取图像的有效特征,从而有效地解决图像识别问题. 针对这一难点,本文主要在残差网络(ResNet)的基础上引入空间变换网络. 空间变换网络可以有效地提取目标区域特征,提高图像识别效率. 同时由于Softmax分类器提取的特征区分并不明显,甚至存在类内间距大于类间间距弊端. 但在图像识别任务中期望特征不仅可分,而且要求类间分别提取的特征区分差异大. 针对这一问题,本文在软最大值(Softmax)分类器中引入中心损失函数(Center Loss). Center Loss损失函数能够使得提取的特征类间距离大,类内距离小,从而提高提取的特征识别度. 在公开的CIFAR10数据集上,该模型取得了不错的性能,识别准确率达到了89%. 相同实验条件下,相对于未改善的残差网络模型,本文提出的模型在公开的CIFAR10数据集识别正确率提高了6%.
Abstract:
With the development of big data era,deep learning has gradually become more practical,leading the development of the era of artificial intelligence. Convolution neural network plays a very important role in image recognition,and it is one of the important components of deep learning model.The key point of image recognition is how to extract the effective features of the image,so as to effectively solve the problem of image recognition. In view of this difficulty,the main work of this paper is to introduce spatial transformation network on the basis of residual network(ResNet). The spatial transformation network can effectively extract the region of interest and improve the efficiency of image recognition. At the same time for the feature extracted by Softmax classifier is not good. In many cases,the intra-class spacing is even larger than the inter-class spacing,but in the image recognition task,the expected features are not only divisible,and with require great differences. In order to solve this problem,this paper introduces the Center Loss function into the Softmax classifier. Center Loss function can make the distance between the extracted feature classes larger and the intra-class distance smaller,thus improving the recognition degree of the extracted features. In the open CIFAR10 dataset,the model has achieved good performance,and the correct recognition rate is up to 89%. Under the same experimental conditions,compared with the unmodified residual network model,the proposed model improves the recognition accuracy of open CIFAR10 dataset by 6%.

参考文献/References:

[1] LECUN Y,BOTTOU L,BENGIO Y,et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE,1998,86(11):2278-2324.
[2]HE K,ZHANG X,REN S,et al. Deep residual learning for image recognition[C]//IEEE Conference on Computer Vision & Pattern Recognition. USA:IEEE,2016.
[3]刘万军,梁雪剑,曲海成. 自适应增强卷积神经网络图像识别[J]. 中国图象图形学报,2017,22(12):1723-1736.
[4]曾维亮,林志贤,陈永洒. 基于卷积神经网络的智能冰箱果蔬图像识别的研究[J]. 微型机与应用,2017,36(8):56-59.
[5]REN S,HE K,GIRSHICK R,et al. Faster R-CNN:towards real-time object detection with region proposal networks[J]. IEEE transactions on pattern analysis and machine intelligence,2017,39(6):1137-1149.
[6]LIU M,SHI J,LI Z,et al. Towards better analysis of deep convolutional neural networks[J]. IEEE transactions on visualization and computer graphics,2017,23(1):91-100.
[7]SILVER D,SCHRITTWIESER J,SIMONYAN K,et al. Mastering the game of go without human knowledge[J]. Nature,2017,550(7676):354.
[8]ZHU P,ISAACS J,BO F,et al. Deep learning feature extraction for target recognition and classification in underwater sonar images[C]//2017 IEEE 56th Annual Conference on Decision and Control(CDC). Melbourne,Australia:IEEE,2018.
[9]WEI W,ZHAO M,WANG J. Effective android malware detection with a hybrid model based on deep autoencoder and convolutional neural network[J]. Journal of ambient intelligence & humanized computing,2018(1):1-9.
[10]刘晨,曲长文,周强,等. 基于卷积神经网络迁移学习的SAR图像目标分类[J]. 现代雷达,2018(3):38-42.
[11]WANG Y,FEI L,ZHANG K,et al. LFNet:a novel bidirectional recurrent convolutional neural network for light-field image super-resolution[J]. IEEE transactions on image processing,2018,27(9):19-26.
[12]CHEN B,ZHANG Y,XIN P. An effective time-domain microwave image reconstruction algorithm for loss-y layered media utilizing the ADI-FDTD method[J]. Journal of China universities of posts & telecommunications,2018,25(2):93-99.
[13]MALLAHI M E,ZOUHRI A,QJIDAA H. Radial meixner moment invariants for 2D and 3D image recognition[J]. International journal of automation & computing,2018,28(2):207-216.
[14]ARCOS G A ?,?LVAREZ G A J A,SORIA M L M. Deep neural network for traffic sign recognition systems:an analysis of spatial transformers and stochastic optimisation methods[J]. Neural Netw,2018,99(12):158-165.
[15]JIN X B,XIE G S,HUANG K,et al. Discriminant zero-shot learning with center loss[J]. Cognitive computation,2019(7):1-10.
[16]曲之琳,胡晓飞. 基于改进激活函数的卷积神经网络研究[J]. 计算机技术与发展,2017(12):83-86.

相似文献/References:

[1]朱志宾,丁世飞.基于TWSVM的图像分类[J].南京师范大学学报(自然科学版),2014,37(03):8.
　Zhu Zhibin,Ding Shifei.Image Classification Based on Twin Support Vector Machines[J].Journal of Nanjing Normal University(Natural Science Edition),2014,37(03):8.
[2]舒 速,杨 明,赵振凯.基于分水岭的高光谱图像分类方法[J].南京师范大学学报(自然科学版),2015,38(01):91.
　Shu Su,Yang Ming,Zhao Zhenkai.Hyperspectral Image Classification Method Based on Watershed[J].Journal of Nanjing Normal University(Natural Science Edition),2015,38(03):91.
[3]郑德鹏,杜吉祥,翟传敏.基于深度学习MPCANet的年龄估计[J].南京师范大学学报(自然科学版),2017,40(01):20.[doi:10.3969/j.issn.1001-4616.2017.01.004]
　Zheng Depeng,Du Jixiang,Zhai Chuanmin.Age Estimation Based on Deep Learning MPCANet[J].Journal of Nanjing Normal University(Natural Science Edition),2017,40(03):20.[doi:10.3969/j.issn.1001-4616.2017.01.004]
[4]朱 繁,王洪元,张 继.基于深度学习的行人重识别研究综述[J].南京师范大学学报(自然科学版),2018,41(04):93.[doi:10.3969/j.issn.1001-4616.2018.04.015]
　Zhu Fan,Wang Hongyuan,Zhang Ji.A Survey of Person Re-identification Based on Deep Learning[J].Journal of Nanjing Normal University(Natural Science Edition),2018,41(03):93.[doi:10.3969/j.issn.1001-4616.2018.04.015]
[5]王 芃,吕 静,沈华乐.基于局部结构保持的自适应有序回归学习[J].南京师范大学学报(自然科学版),2019,42(02):9.[doi:10.3969/j.issn.1001-4616.2019.02.002]
　Wang Peng,Lü Jing,Shen Huale.Improved Adaptive Ordinal Regression LearningBased on Locality Structure Preserving[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):9.[doi:10.3969/j.issn.1001-4616.2019.02.002]
[6]孙茹君,张鲁飞.基于动态指导的深度学习模型稀疏化执行方法[J].南京师范大学学报(自然科学版),2019,42(03):11.[doi:10.3969/j.issn.1001-4616.2019.03.002]
　Sun Rujun,Zhang Lufei.Dynamic Sparse Method for Deep Learning Execution[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):11.[doi:10.3969/j.issn.1001-4616.2019.03.002]
[7]赵文芳,林润生,唐 伟,等.基于深度学习的PM2.5短期预测模型[J].南京师范大学学报(自然科学版),2019,42(03):32.[doi:10.3969/j.issn.1001-4616.2019.03.005]
　Zhao Wenfang,Lin Runsheng,Tang Wei,et al.Forecasting Model of Short-Term PM2.5 ConcentrationBased on Deep Learning[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):32.[doi:10.3969/j.issn.1001-4616.2019.03.005]
[8]张新峰,闫昆鹏,赵 珣.基于双向LSTM的手写文字识别技术研究[J].南京师范大学学报(自然科学版),2019,42(03):58.[doi:10.3969/j.issn.1001-4616.2019.03.008]
　Zhang Xinfeng,Yan Kunpeng,Zhao Xun.Handwriting Chinese Text Recognition Using BiLSTM Network[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):58.[doi:10.3969/j.issn.1001-4616.2019.03.008]
[9]贾玉福,胡胜红,刘文平,等.使用条件生成对抗网络的自然图像增强方法[J].南京师范大学学报(自然科学版),2019,42(03):88.[doi:10.3969/j.issn.1001-4616.2019.03.012]
　Jia Yufu,Hu Shenghong,Liu Wenping,et al.Wild Image Enhancement with Conditional Generative Adversarial Network[J].Journal of Nanjing Normal University(Natural Science Edition),2019,42(03):88.[doi:10.3969/j.issn.1001-4616.2019.03.012]
[10]汪 晨,张辉辉,乐继旺,等.基于深度学习和遥感影像的松材线虫病疫松树目标检测[J].南京师范大学学报(自然科学版),2021,44(03):84.[doi:10.3969/j.issn.1001-4616.2021.03.013]
　Wang Chen,Zhang Huihui,Le Jiwang,et al.Object Detection to the Pine Trees Affected by Pine Wilt Diseasein Remote Sensing Images Using Deep Learning[J].Journal of Nanjing Normal University(Natural Science Edition),2021,44(03):84.[doi:10.3969/j.issn.1001-4616.2021.03.013]

备注/Memo

备注/Memo:
收稿日期:2019-07-05.基金项目:块数据中多源异构条数据关联识别理论模型研究、贵州省公共大数据重点实验室开放课题(2017BDKFJJ034). 通讯联系人:王旭,博士,副教授,研究方向:大数据应用、人工智能、量子通讯. E-mail:xuwang@gzu.edu.cn
更新日期/Last Update: 2019-09-30