基于生成对抗网络的零样本图像分类

魏宏喜; 张越

doi:10.13700/j.bh.1001-5965.2019.0363

基于生成对抗网络的零样本图像分类

doi: 10.13700/j.bh.1001-5965.2019.0363

魏宏喜^,,
张越

内蒙古大学计算机学院, 呼和浩特 010021

基金项目:

国家自然科学基金 61463038

内蒙古自治区自然科学基金重大项目 2019ZD14

详细信息

作者简介:
魏宏喜男, 博士, 教授, 博士生导师。主要研究方向:机器学习与数据挖掘

张越男, 硕士研究生。主要研究方向:图像分类

通讯作者:
魏宏喜. E-mail: cswhx@imu.edu.cn

中图分类号: TP391
计量
- 文章访问数: 1297
- HTML全文浏览量: 120
- PDF下载量: 529
- 被引次数: 0
出版历程
- 收稿日期: 2019-07-08
- 录用日期: 2019-08-03
- 网络出版日期: 2019-12-20

Zero-shot image classification based on generative adversarial network

WEI Hongxi^,,
ZHANG Yue

College of Computer Science, Inner Mongolia University, Hohhot 010021, China

Funds:

National Natural Science Foundation of China 61463038

Natural Science Foundation of Inner Mongolia Autonomous Region 2019ZD14

More Information

Corresponding author: WEI Hongxi. E-mail: cswhx@imu.edu.cn

摘要

摘要:
在图像分类任务中，零样本图像分类问题已成为一个研究热点。为了解决零样本图像分类问题，采用一种基于生成对抗网络（GAN）的方法，通过生成未知类的图像特征使得零样本分类任务转换为传统的图像分类任务。同时对生成对抗网络中的判别网络做出改进，使其判别过程更加准确，从而进一步提高生成图像特征的质量。实验结果表明：所提方法在AWA、CUB和SUN数据集上的分类准确率分别提高了0.4%、0.4%和0.5%。因此，所提方法通过改进生成对抗网络，能够生成质量更好的图像特征，从而有效解决零样本图像分类问题。
- 深度学习 /
- 图像分类 /
- 零样本学习 /
- 生成对抗网络(GAN) /
- 图像特征生成
Abstract:
The problem of zero-shot image classification has become a research focus in the field of image classification. In this paper, a method based on generative adversarial network (GAN) is used to solve the problem of zero-shot image classification. By generating image features of unseen classes, the zero-shot classification task is transformed into a conventional image classification task. At the same time, this paper makes modifications to the discriminant network in the generative adversarial network to make the discriminating process more accurate. The experimental results show that the performance of the proposed method has been increased by 0.4%, 0.4% and 0.5% on the datasets of AWA, CUB and SUN, respectively. Therefore, the proposed method can generate the better features by improving the generative adversarial networks, which results in solving the problem of zero-shot image calssification effectively.
- deep learning /
- image classification /
- zero-shot learning /
- generative adversarial network (GAN) /
- image feature generation

HTML全文

图 1 零样本学习示例

Figure 1. An example of zero-shot learning

下载: 全尺寸图片幻灯片

图 2 f-CLWSGAN模型的网络结构

Figure 2. Network structure of f-CLSWGAN model

下载: 全尺寸图片幻灯片

图 3 FD-fGAN模型的网络结构

Figure 3. Network structure of FD-fGAN model

下载: 全尺寸图片幻灯片

图 4 不同数据集上的分类准确率

Figure 4. Classification accuracy on different datasets

下载: 全尺寸图片幻灯片

表 1 三个数据集的划分情况

Table 1. Division of three datasets

数据集	类别数	已知类别数	未知类别数
AWA	50	40	10
CUB	200	150	50
SUN	717	645	72

下载: 导出CSV

表 2 不同方法在3个数据集上的分类准确率比较

Table 2. Comparison of classification accuracy of different methods on three datasets

方法	分类准确率/%
方法	AWA	CUB	SUN
DAP^[6]	44.1	40.4	39.9
CONSE^[16]	45.6	34.3	38.8
SSE^[17]	60.1	43.9	51.5
DeViSE^[18]	54.2	52.0	56.5
SJE^[19]	65.6	53.9	53.7
ESZSL^[7]	58.2	53.9	54.5
ALE^[20]	59.9	54.9	58.1
SYNC^[2]	54.0	55.6	56.3
f-CLSWGAN^[9]	68.2	57.3	60.8
FD-fGAN	68.6	57.7	61.3

下载: 导出CSV

参考文献(20)

[1]	LAROCHELLE H, ERHAN D, BENGIO Y.Zero-data learning of new tasks[C]//Proceedings of the Association for the Advancement of Artificial Intelligence. Palo Alto: AAAI Press, 2008: 646-651. http://www.cs.toronto.edu/~larocheh/publications/aaai2008_zero-data.pdf
[2]	CHANGPINYO S, CHAO W, GONG B, et al.Synthesized classifiers for zero-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2016: 5327-5336. https://www.cv-foundation.org/openaccess/content_cvpr_2016/papers/Changpinyo_Synthesized_Classifiers_for_CVPR_2016_paper.pdf
[3]	KODIROV E, XIANG T, GONG S.Semantic auto-encoder for zero-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2017: 4447-4456. http://openaccess.thecvf.com/content_cvpr_2017/papers/Kodirov_Semantic_Autoencoder_for_CVPR_2017_paper.pdf
[4]	XIAN Y, SCHIELE B, AKATA Z.Zero-shot learning-the good, the bad and the ugly[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2017: 3077-3086. https://arxiv.org/pdf/1703.04394.pdf
[5]	DING Z, SHAO M, FU Y.Low-rank embedded ensemble semantic dictionary for zero-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2017: 6005-6013. http://openaccess.thecvf.com/content_cvpr_2017/papers/Ding_Low-Rank_Embedded_Ensemble_CVPR_2017_paper.pdf
[6]	LAMPERT C H, NICKISCH H, HARMELING S.Attribute-based classification for zero-shot visual object categorization[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(3):453-465. doi: 10.1109/TPAMI.2013.140
[7]	ROMERA-PAREDES B, TORR P.An embarrassingly simple approach to zero-shot learning[C]//Proceedings of the International Conference on Machine Learning.Madison: Omnipress, 2015: 2152-2161. http://proceedings.mlr.press/v37/romera-paredes15.pdf
[8]	GOODFELLOW I, POUGET-ABADIE J, MIRZA M, et al.Generative adversarial nets[C]//Proceedings of the Neural Information Processing Systems.Cambridge: MIT Press, 2014: 2672-2680. https://arxiv.org/pdf/1406.2661v1.pdf
[9]	XIAN Y, LORENZ T, SCHIELE B, et al.Feature generating networks for zero-shot learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2018: 5542-5551. https://arxiv.org/pdf/1712.00981.pdf
[10]	ZHU Y, ELHOSEINY M, LIU B, et al.A generative adversarial approach for zero-shot learning from noisy texts[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2018: 1004-1013. http://openaccess.thecvf.com/content_cvpr_2018/papers/Zhu_A_Generative_Adversarial_CVPR_2018_paper.pdf
[11]	ARJOVSKY M, CHINTALA S, BOTTOU L.Wasserstein generative adversarial networks[C]//Proceedings of the International Conference on Machine Learning.Madison: Omnipress, 2017: 214-223.
[12]	MIRZA M, OSINDERO S.Conditional generative adversarial nets[EB/OL].(2014-11-06)[2019-07-01].https://arxiv.org/abs/1411.1784.
[13]	LAMPERT C H, NICKISCH H, HARMELING S.Learning to detect unseen object classes by between-class attribute transfer[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2009: 951-958. https://blog.csdn.net/u011070272/article/details/73250102
[14]	WAH C, BRANSON S, WELINDER P, et al.The Caltech-UCSD Birds-200-2011 dataset:CNS-TR-2011-001[R].Pasadena:California Institute of Technology, 2011.
[15]	PATTERSON G, HAYS J.SUN attribute database: Discovering, annotating, and recognizing scene attributes[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2012: 2751-2758.
[16]	NOROUZI M, MIKOLOV T, BENGIO S.Zero-shot learning by convex combination of semantic embeddings[C]//Proceedings of the International Conference on Learning Representations.Brookline: Microtome Publishing, 2014: 10-19. https://arxiv.org/abs/1312.5650
[17]	ZHANG Z, SALIGRAMA V.Zero-shot learning via semantic similarity embedding[C]//Proceedings of the IEEE International Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2015: 4166-4174. https://arxiv.org/pdf/1509.04767.pdf
[18]	FROME A, CORRADO G S, SHLENS J, et al.DeViSE: A deep visual semantic embedding model[C]//Proceedings of the Neural Information Processing Systems.Cambridge: MIT Press, 2013: 2121-2129. http://www.cs.toronto.edu/~ranzato/publications/frome_nips2013.pdf
[19]	AKATA Z, REED S, WALTER D, et al.Evaluation of output embeddings for fine-grained image classification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.Piscataway, NJ: IEEE Press, 2015: 2927-2936. https://www.cv-foundation.org/openaccess/content_cvpr_2015/papers/Akata_Evaluation_of_Output_2015_CVPR_paper.pdf
[20]	AKATA Z, PERRONNIN F, HARCHAOUI Z, et al.Label-embedding for image classification[J].IEEE Transactions on Pattern Analysis and Machine Intelligence, 2016, 38(7):1425-1438. doi: 10.1109/TPAMI.2015.2487986