Advanced search
1 file | 2.94 MB Add to list

Deep feature fusion through adaptive discriminative metric learning for scene recognition

Chen Wang (UGent) , Guohua Peng and Bernard De Baets (UGent)
(2020) INFORMATION FUSION. 63. p.1-12
Author
Organization
Abstract
With the development of deep learning techniques, fusion of deep features has demonstrated the powerful capability to improve recognition performance. However, most researchers directly fuse different deep feature vectors without considering the complementary and consistent information among them. In this paper, from the viewpoint of metric learning, we propose a novel deep feature fusion method, called deep feature fusion through adaptive discriminative metric learning (DFF-ADML), to explore the complementary and consistent information for scene recognition. Concretely, we formulate an adaptive discriminative metric learning problem, which not only fully exploits discriminative information from each deep feature vector, but also adaptively fuses complementary information from different deep feature vectors. Besides, we map different deep feature vectors of the same image into a common space by different linear transformations, such that the consistent information can be preserved as much as possible. Moreover, DFF-ADML is extended to a kernelized version. Extensive experiments on both natural scene and remote sensing scene datasets demonstrate the superiority and robustness of the proposed deep feature fusion method.
Keywords
Signal Processing, Hardware and Architecture, Software, Information Systems, Deep feature fusion, Adaptive discriminative metric learning, Scene recognition, IMAGE CLASSIFICATION, SCALE

Downloads

  • (...).pdf
    • full text (Published version)
    • |
    • UGent only
    • |
    • PDF
    • |
    • 2.94 MB

Citation

Please use this url to cite or link to this publication:

MLA
Wang, Chen, et al. “Deep Feature Fusion through Adaptive Discriminative Metric Learning for Scene Recognition.” INFORMATION FUSION, vol. 63, 2020, pp. 1–12, doi:10.1016/j.inffus.2020.05.005.
APA
Wang, C., Peng, G., & De Baets, B. (2020). Deep feature fusion through adaptive discriminative metric learning for scene recognition. INFORMATION FUSION, 63, 1–12. https://doi.org/10.1016/j.inffus.2020.05.005
Chicago author-date
Wang, Chen, Guohua Peng, and Bernard De Baets. 2020. “Deep Feature Fusion through Adaptive Discriminative Metric Learning for Scene Recognition.” INFORMATION FUSION 63: 1–12. https://doi.org/10.1016/j.inffus.2020.05.005.
Chicago author-date (all authors)
Wang, Chen, Guohua Peng, and Bernard De Baets. 2020. “Deep Feature Fusion through Adaptive Discriminative Metric Learning for Scene Recognition.” INFORMATION FUSION 63: 1–12. doi:10.1016/j.inffus.2020.05.005.
Vancouver
1.
Wang C, Peng G, De Baets B. Deep feature fusion through adaptive discriminative metric learning for scene recognition. INFORMATION FUSION. 2020;63:1–12.
IEEE
[1]
C. Wang, G. Peng, and B. De Baets, “Deep feature fusion through adaptive discriminative metric learning for scene recognition,” INFORMATION FUSION, vol. 63, pp. 1–12, 2020.
@article{8663720,
  abstract     = {{With the development of deep learning techniques, fusion of deep features has demonstrated the powerful capability to improve recognition performance. However, most researchers directly fuse different deep feature vectors without considering the complementary and consistent information among them. In this paper, from the viewpoint of metric learning, we propose a novel deep feature fusion method, called deep feature fusion through adaptive discriminative metric learning (DFF-ADML), to explore the complementary and consistent information for scene recognition. Concretely, we formulate an adaptive discriminative metric learning problem, which not only fully exploits discriminative information from each deep feature vector, but also adaptively fuses complementary information from different deep feature vectors. Besides, we map different deep feature vectors of the same image into a common space by different linear transformations, such that the consistent information can be preserved as much as possible. Moreover, DFF-ADML is extended to a kernelized version. Extensive experiments on both natural scene and remote sensing scene datasets demonstrate the superiority and robustness of the proposed deep feature fusion method.}},
  author       = {{Wang, Chen and Peng, Guohua and De Baets, Bernard}},
  issn         = {{1566-2535}},
  journal      = {{INFORMATION FUSION}},
  keywords     = {{Signal Processing,Hardware and Architecture,Software,Information Systems,Deep feature fusion,Adaptive discriminative metric learning,Scene recognition,IMAGE CLASSIFICATION,SCALE}},
  language     = {{eng}},
  pages        = {{1--12}},
  title        = {{Deep feature fusion through adaptive discriminative metric learning for scene recognition}},
  url          = {{http://dx.doi.org/10.1016/j.inffus.2020.05.005}},
  volume       = {{63}},
  year         = {{2020}},
}

Altmetric
View in Altmetric
Web of Science
Times cited: