基于共享特征相對(duì)屬性的零樣本圖像分類
doi: 10.11999/JEIT161133
-
1.
(中國(guó)科學(xué)院電子學(xué)研究所蘇州研究院 蘇州 215123) ②(中國(guó)科學(xué)技術(shù)大學(xué)軟件學(xué)院 合肥 231000)
基金項(xiàng)目:
國(guó)家自然科學(xué)基金(41501485)
Shared Features Based Relative Attributes forZero-shot Image Classification
-
1.
(Suzhou Institute, Institute of Electronics, Chinese Academy of Sciences, Suzhou 215123, China)
-
2.
(School of Software Engineering, University of Science and Technology of China, Hefei 231000, China)
Funds:
The National Natural Science Foundation of China (41501485)
-
摘要: 在利用相對(duì)屬性學(xué)習(xí)實(shí)現(xiàn)零樣本圖像分類中，現(xiàn)有的方法并沒有考慮屬性與類別之間的關(guān)系，為此該文提出一種基于共享特征相對(duì)屬性的零樣本圖像分類方法。該方法采用多任務(wù)學(xué)習(xí)的思想來共同學(xué)習(xí)類別分類器和屬性分類器，獲得一個(gè)低維的共享特征子空間，挖掘?qū)傩耘c類別之間的關(guān)系。同時(shí)，利用共享特征來學(xué)習(xí)屬性排序函數(shù)，得到基于共享特征的相對(duì)屬性模型，解決了相對(duì)屬性學(xué)習(xí)過程中丟失屬性與類別關(guān)系的問題。另外，將基于共享特征的相對(duì)屬性模型用于零樣本圖像分類中，有效提高了零樣本圖像分類的識(shí)別率。實(shí)驗(yàn)數(shù)據(jù)集上的結(jié)果表明，該方法具有較高的相對(duì)屬性學(xué)習(xí)性能和零樣本圖像分類精度。
-
關(guān)鍵詞:
- 相對(duì)屬性 /
- 多任務(wù)學(xué)習(xí) /
- 共享特征 /
- 零樣本圖像分類
Abstract: Most algorithms of the zero-shot image classification with relative attributes do not consider the relationship between attributes and classes, therefore a new relative attributes method based on shared features is proposed for zero-shot image classification. In analogy to the multi-task learning, the object classifier and attribute classifier are simultaneously learned in this method, from which a shared sub-space of lower dimensional features is obtained to mine the relationship between attributes and classes. Inspired by the success of shared features, a novel relative attributes model based on shared features is proposed to promote the performance of the relationship between attributes and classes, in which the ranking function per attribute is learned by using shared features. In addition, the novel relative attributes model based on shared features is applied to zero-shot image classification, which yields high accuracy due to the shared features included. Experimental results demonstrate that the proposed method can achieve high relative attributes learning efficiency and zero-shot image classification accuracy. -
GAN C, YANG T, and GONG B. Learning attributes equals multi-source domain generalization[C]. IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, NV, USA, 2016: 87-97. QIN Jie, WANG Yunhong, LIU Li, et al. Beyond semantic attributes: Discrete latent attributes learning for zero-shot recognition[J]. IEEE Signal Processing Letters, 2016, 23(11): 1667-1671. doi: 10.1109/LSP.2016.2612247. PARIKH D and GRAUMAN K. Relative attributes[C]. IEEE International Conference on Computer Vision, Barcelona, Spain, 2011: 503-510. YANG X, ZHANG T, XU C, et al. Deep relative attributes[J]. IEEE Transactions on Multimedia, 2016, 18(9): 1832-1842. doi: 10.1109/TMM.2016.2582379. CHEN L, ZHANG Q, and LI B X. Predicting multiple attributes via relative multi-task learning[C]. IEEE Conference on Computer Vision and Pattern Recognition, Columbus, OH, USA, 2014: 1027-1034. 和華, 杜蘭, 徐丹蕾, 等. 基于多任務(wù)復(fù)數(shù)因子分析模型的雷達(dá)高分辨距離像識(shí)別方法[J]. 電子與信息學(xué)報(bào), 2015, 37(10): 2307-2313. doi: 10.11999/JEIT141591. HE Hua, DU Lan, XU Danlei, et al. Radar HRRP target recognition method based on multi-task learning and complex factor analysis[J]. Journal of Electronics Information Technology, 2015, 37(10): 2307-2313. doi: 10.11999/JEIT141591. HWANG S J, SHA F, and GRAUMAN K. Sharing features between objects and their attributes[C]. IEEE Conference on Computer Vision and Pattern Recognition, Colorado Springs, CO, USA, 2011: 1761-1768. LAMPERT C H, NICKISCH H, and HARMELING S. Attribute-based classification for zero-shot visual object categorization[J]. Pattern Analysis and Machine Intelligence, 2014, 36(3): 453-465. doi: 10.1109/TPAMI.2013.140. ARGYRIOU A, EVGENIOU T, and PONTIL M. Convex multi-task feature learning[J]. Machine Learning, 2008, 73(3): 243-272. doi: 10.1007/s10994-007-5040-8. SHI C, RUAN Q, AN G, et al. Hessian semi-supervised sparse feature selection based on L2,1/2-matrix norm[J]. IEEE Transactions on Multimedia, 2015, 17(1): 16-28. doi:10.1109/ TMM.2014.2375792. 李秀友, 薛永華, 董云龍, 等. 基于迭代凸優(yōu)化的恒模波形合成方法[J]. 電子與信息學(xué)報(bào), 2015, 37(9): 2171-2176. doi: 10.11999/JEIT141593. LI Xiuyou, XUE Yonghua, DONG Yunlong, et al. Constant modulus waveform synthesis based on iterative convex optimization[J]. Journal of Electronics Information Technology, 2015, 37(9): 2171-2176. doi: 10.11999/ JEIT141593. YANG Z M, WU H J, LI C N, et al. Least squares recursive projection twin support vector machine for multi-class classification[J]. International Journal of Machine Learning Cybernetics, 2016, 7(3): 1-16. doi: 10.1007/s13042-015- 0394-x. 及歆榮, 侯翠琴, 侯義斌. 無線傳感器網(wǎng)絡(luò)下線性支持向量機(jī)分布式協(xié)同訓(xùn)練方法研究[J]. 電子與信息學(xué)報(bào), 2015(3): 708-714. doi: 10.11999/JEIT140408 JI Xinrong, HOU Cuiqin, and HOU Yibin. Research on the distributed training method for linear SVM in WSN[J]. Journal of Electronics Information Technology, 2015, 37(3): 708-714. doi: 10.11999/JEIT140408. LI S, SHAN S, and CHEN X. Relative forest for attribute prediction[C]. Asian Conference on Computer Vision, Daejeon, Korea, 2012: 316-327. JAYARAMAN D and GRAUMAN K. Zero shot recognition with unreliable attributes[C]. Conference on Neural Information Processing Systems, Montreal, QC, Canada, 2014: 3464-3472. HUANG S, ELHOSEINY M, ELGAMMAL A, et al.. Learning hypergraph-regularized attribute predictors[C]. IEEE Conference on Computer Vision and Pattern Recognition, Boston, MA, USA, 2015: 409-417. XUE J H and HALL P. Why does rebalancing class-unbalanced data improve AUC for linear discriminant analysis?[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2015, 37(5): 1109-12. doi: 10.1109/ TPAMI.2014.2359660. -
計(jì)量
- 文章訪問數(shù): 1380
- HTML全文瀏覽量: 259
- PDF下載量: 420
- 被引次數(shù): 0
下載:
