[1]经卓勋,刘建明*.面向细粒度分类的预测属性引导的注意力研究[J].江西师范大学学报(自然科学版),2022,(04):379-385.[doi:10.16357/j.cnki.issn1000-5862.2022.04.08]
 JING Zhuoxun,LIU Jianming*.The Method on Fine-Grained Image Categorization Using Predicted-Attribute Guided Channel Attention Module[J].Journal of Jiangxi Normal University:Natural Science Edition,2022,(04):379-385.[doi:10.16357/j.cnki.issn1000-5862.2022.04.08]
点击复制

面向细粒度分类的预测属性引导的注意力研究()
分享到:

《江西师范大学学报》(自然科学版)[ISSN:1006-6977/CN:61-1281/TN]

卷:
期数:
2022年04期
页码:
379-385
栏目:
信息科学与技术
出版日期:
2022-07-25

文章信息/Info

Title:
The Method on Fine-Grained Image Categorization Using Predicted-Attribute Guided Channel Attention Module
文章编号:
1000-5862(2022)04-0379-07
作者:
经卓勋刘建明*
江西师范大学计算机信息工程学院,江西 南昌 330022
Author(s):
JING ZhuoxunLIU Jianming*
School of Computer Information Engineering,Jiangxi Normal University,Nanchang Jiangxi 330022,China
关键词:
细粒度图像分类 注意力机制 属性预测
Keywords:
fine grained image classification attention n mechanism attribute prediction
分类号:
TP 311
DOI:
10.16357/j.cnki.issn1000-5862.2022.04.08
文献标志码:
A
摘要:
细粒度图像分类任务比一般图像分类任务更具有挑战性,其通常需要对类间差异小、类内差异大的样本进行分类.现有细粒度分类方法主要依赖视觉特征进行分类,而人类可以根据文本描述等属性描述来辅助识别图像类别.该文提出了一种通过预测属性引导的通道注意力模块,该模块可以插入到任意的卷积神经网络中,从而让模型学习到更高级的特征表示.最后,该算法在CUB-200-2011数据集上测试,在使用Resnet-50、VGG-19、Bilinear-CNN作为主干网络训练时的精度分别达到87.1%、82.1%、85.5%,精度得到显著提升.
Abstract:
The fine-grained image classification task is more challenging than the general image classification task.Its task usually needs to classify the samples with low inter-class but high intra-class variation.The existing fine-grained classification methods mainly rely on visual features for classification,but human beings can recognize image categories according to text attribute description.To this end,the predicted-attribute guided channel attention module is proposed.The module can insert any convolutional neural network,which is intended to make the model to learn more advanced feature representation.Finally,the algorithm proposed in this paper tests on CUB-200-2011 dataset.The algorithm achieves 87.1%,82.1%,85.5% when training using Resnet-50,VGG-19,Bilinear-CNN as backbone network.It can observe a significant improvement in accuracy.

参考文献/References:

[1] SONG Kaitao,WEI Xiushen,SHU Xiangbo,et al.Bi-modal progressive mask attention for fine-grained recognition[J].IEEE Transactions on Image Processing,2020,29:7006-7018.
[2] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[J].Communications of the ACM,2017,60(6):84-90.
[3] SHI Xingjian,CHEN Zhourong,WANG Hao,et al.Convolutional LSTM network:a machine learning approach for precipitation nowcasting[EB/OL].[2021-12-17].https://arxiv.org/pdf/1506.042.pdf.
[4] HE Xiangteng,PENG Yuxin.Fine-grained image classification via combining vision and language[EB/OL].[2021-11-13].https://arxiv.org/abs/1704.02792v1.
[5] GAO Yu,HAN Xintong,WANG Xun,et al.Channel interaction networks for fine-grained image categorization[J].AAAI-20 Technical Track 7,2020,34(7):10818-10825.
[6] ZHUANG Peiqin,WANG Yali,QIAO Yu.Learning attentive pairwise interaction for fine-grained classification[J].AAAI-20 Technical Track 7,2020,34(7):13130-13137.
[7] HU Tao,Qi Honggang.See better before looking closer:weakly supervised data augmentation network for fine-grained visual classification[EB/OL].[2021-11-12].https://arxiv.org/abs/1901.09891v1.
[8] ZHENG Heliang,FU Jianlong,ZHA Zhengjun,et al.Looking for the devil in the details:learning trilinear attention sampling network for fine-grained image recognition[EB/OL].[2021-11-15].https://arxiv.org/abs/1903.06150v2.
[9] DU Ruoyi,CHANG Dongliang,BHUNIA A K,et al.Fine-grained visual classification via progressive multi-granularity training of jigsaw patches[EB/OL].[2021-10-19].https://arxiv.org/abs/2003.03836v2.
[10] DING Yifeng,MA Zhanyu,WEN Shaoguo,et al.AP-CNN:weakly supervised attention pyramid convolutional neural network for fine-grained visual classification[J].IEEE Transactions on Image Processing,2021,30:2826-2836.
[11] BEHERA A,WHARTON Z,HEWAGE P,et al.Context-aware attentional pooling(cap)for fine-grained visual classification[EB/OL].[2021-10-13].https://doi.org/10.48550/arXiv.2101.06635.
[12] JI Ruyi,WEN Longyin,ZHANG Libo,et al.Attention convolutional binary neural tree for fine-grained visual categorization[EB/OL].[2021-11-10].https://ieeexplore.ieee.org/document/9157539.
[13] HE Ju,CHEN Jieneng,LIU Shuai,et al.TransfG:a transformer architecture for fine-grained recognition[EB/OL].[2021-12-03].https://doi.org/10.48550/arXiv.2103.07976.
[14] CHEN Tianshui,LIN Liang,CHEN Riquan,et al.Knowledge-embedded representation learning for fine-grained image recognition[EB/OL].[2021-12-09].https://doi.org/10.48550/arXiv.1807.00505.
[15] SUN Liang,GUAN Xiang,YANG Yang,et al.Text-embedded bilinear model for fine-grained visual recognition[EB/OL].[2021-10-13].https://doi.org/10.1145/3394171.3413638.
[16] ZHU Linchao,YANG Yi.ActBERT:learning global-local video-text representations[EB/OL].[2021-10-18].https://doi.org/10.48550/arXiv.2011.07231.
[17] MNIH V,HEESS N,GRAVES A.Recurrent models of visual attention[EB/OL].[2021-10-18].http://de.arxiv.org/pdf/1406.6247.
[18] WOO S,PARK J,LEE J Y,et al.CBAM:convolutional block attention module[EB/OL].[2021-10-16].https://arxiv.org/pdf/1807.06521.pdf.
[19] PARK J,WOO S,LEE J Y,et al.BAM:bottleneck attention module[EB/OL].[2021-11-13].https://arxiv.org/pdf/1807.06514.pdf
[20] WAH C,BRANSON S,WELINDER P,et al.The CALTECH-UCSD birds-200-2011 dataset[EB/OL].[2021-11-16]. https://citeseerx.ist.psu.edu/viewdoc/download; jsessionid=FF4D49EB69FF280011EEF1AA90D8050A?doi=10.1.1.372.852&rep=rep1&type=pdf.

相似文献/References:

[1]何金宝,胡秋宝,付志超,等.基于DCGAN和改进YOLOv5s的桥梁表面缺陷检测识别[J].江西师范大学学报(自然科学版),2022,(06):655.[doi:10.16357/j.cnki.issn1000-5862.2022.06.14]
 HE Jinbao,HU Qiubao,FU Zhichao,et al.The Bridge Apparent Defects Detection Based on DCGAN and Improved YOLOv5s[J].Journal of Jiangxi Normal University:Natural Science Edition,2022,(04):655.[doi:10.16357/j.cnki.issn1000-5862.2022.06.14]

备注/Memo

备注/Memo:
收稿日期:2022-02-18
基金项目:国家自然科学基金(61662034)和江西省省自然科学基金(20202BAB202020)资助项目.
通信作者:刘建明(1981—),男,江西鹰潭人,副教授,博士,主要从事细粒度图像识别、零样本和小样本学习、弱监督学习研究.E-mail:liujianming@jxnu.edu.cn
更新日期/Last Update: 2022-07-25