Kernel Pooling for Convolutional Neural Networks
Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Standard
Kernel Pooling for Convolutional Neural Networks. / Cui, Yin; Zhou, Feng; Wang, Jiang; Liu, Xiao; Lin, Yuanqing; Belongie, Serge.
I: Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017, 2017, s. 3049-3058.Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Kernel Pooling for Convolutional Neural Networks
AU - Cui, Yin
AU - Zhou, Feng
AU - Wang, Jiang
AU - Liu, Xiao
AU - Lin, Yuanqing
AU - Belongie, Serge
N1 - Funding Information: This work was supported in part by Google Focused Re-searchAward,AWS CloudCreditsforResearch, Microsoft Research Award and a Facebook equipment donation. Publisher Copyright: ©2017 IEEE.
PY - 2017
Y1 - 2017
N2 - Convolutional Neural Networks (CNNs) with Bilinear Pooling, initially in their full form and later using compact representations, have yielded impressive performance gains on a wide range of visual tasks, including fine-grained visual categorization, visual question answering, face recognition, and description of texture and style. The key to their success lies in the spatially invariant modeling of pairwise (2nd order) feature interactions. In this work, we propose a general pooling framework that captures higher order interactions of features in the form of kernels. We demonstrate how to approximate kernels such as Gaussian RBF up to a given order using compact explicit feature maps in a parameter-free manner. Combined with CNNs, the composition of the kernel can be learned from data in an endto- end fashion via error back-propagation. The proposed kernel pooling scheme is evaluated in terms of both kernel approximation error and visual recognition accuracy. Experimental evaluations demonstrate state-of-the-art performance on commonly used fine-grained recognition datasets.
AB - Convolutional Neural Networks (CNNs) with Bilinear Pooling, initially in their full form and later using compact representations, have yielded impressive performance gains on a wide range of visual tasks, including fine-grained visual categorization, visual question answering, face recognition, and description of texture and style. The key to their success lies in the spatially invariant modeling of pairwise (2nd order) feature interactions. In this work, we propose a general pooling framework that captures higher order interactions of features in the form of kernels. We demonstrate how to approximate kernels such as Gaussian RBF up to a given order using compact explicit feature maps in a parameter-free manner. Combined with CNNs, the composition of the kernel can be learned from data in an endto- end fashion via error back-propagation. The proposed kernel pooling scheme is evaluated in terms of both kernel approximation error and visual recognition accuracy. Experimental evaluations demonstrate state-of-the-art performance on commonly used fine-grained recognition datasets.
U2 - 10.1109/CVPR.2017.325
DO - 10.1109/CVPR.2017.325
M3 - Conference article
AN - SCOPUS:85044222046
SP - 3049
EP - 3058
JO - Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
JF - Proceedings - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
T2 - 30th IEEE Conference on Computer Vision and Pattern Recognition, CVPR 2017
Y2 - 21 July 2017 through 26 July 2017
ER -
ID: 301826599