logo

SCIENCE CHINA Information Sciences, Volume 61, Issue 10: 109101(2018) https://doi.org/10.1007/s11432-018-9497-0

GlanceNets – efficient convolutional neural networks with adaptive hard example mining

More info
  • ReceivedFeb 27, 2018
  • AcceptedMay 30, 2018
  • PublishedSep 3, 2018

Abstract

There is no abstract available for this article.


Acknowledgment

This work was supported by National Natural Science Foundation of China (Grant No. 61632081) and National Program of Key Basic Research Project (973 Program) (Grant No. 2014CB340400).


References

[1] Huang G, Liu Z, van der Maaten L, et al. Densely connected convolutional networks. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Honolulu, 2017. 2261--2269. Google Scholar

[2] Pang Y W, Sun M L, Jiang X H. Convolution in convolution for network in network. IEEE Trans Neural Netw Learn Syst, 2018, 29: 1587-1597 CrossRef PubMed Google Scholar

[3] Cao J L, Pang Y W, Li X L. Randomly translational activation inspired by the input distributions of ReLU. Neurocomputing, 2018, 275: 859-868 CrossRef Google Scholar

[4] Pang Y W, Cao J L, Li X L. Cascade learning by optimally partitioning. IEEE Trans Cybern, 2017, 47: 4148-4161 CrossRef PubMed Google Scholar

[5] Shrivastava A, Gupta A, Girshick R. Training region-based object detectors with online hard example mining. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 2016. 761--769. Google Scholar

[6] Lin T Y, Goyal P, Girshick R B, et al. Focal loss for dense object detection. In: Proceedings of IEEE Conference on Computer Vision, Venice, 2017. Google Scholar

[7] Teerapittayanon S, McDanel B, Kung H T. BranchyNet: fast inference via early exiting from deep neural networks. In: Proceedings of International Conference on Pattern Recognition, Cancun, 2016. 2464--2469. Google Scholar

[8] Lee C Y, Xie S N, Gallagher P, et al. Deeply-supervised nets. In: Proceedings of International Conference on Artificial Intelligence and Statistics, San Diego, 2015. 562--570. Google Scholar

  • Figure 1

    (Color online) A GlanceNet consists of a backbone CNN architecture and several bypasses, in which the proposed online hard example mining method (green boxes) and threshold learning method (blue boxes) are applied. Detail compositions of a bypass are shown on the right.

Copyright 2019 Science China Press Co., Ltd. 《中国科学》杂志社有限责任公司 版权所有

京ICP备18024590号-1