Corpus ID: 68222714

Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet

@article{Brendel2019ApproximatingCW,
  title={Approximating CNNs with Bag-of-local-Features models works surprisingly well on ImageNet},
  author={Wieland Brendel and Matthias Bethge},
  journal={ArXiv},
  year={2019},
  volume={abs/1904.00760}
}
  • Wieland Brendel, Matthias Bethge
  • Published 2019
  • Computer Science, Mathematics
  • ArXiv
  • Deep Neural Networks (DNNs) excel on many complex perceptual tasks but it has proven notoriously difficult to understand how they reach their decisions. [...] Key Method This strategy is closely related to the bag-of-feature (BoF) models popular before the onset of deep learning and reaches a surprisingly high accuracy on ImageNet (87.6% top-5 for 33 x 33 px features and Alexnet performance for 17 x 17 px features).Expand Abstract

    Figures, Tables, and Topics from this paper.

    Citations

    Publications citing this paper.
    SHOWING 1-10 OF 131 CITATIONS

    Evaluating CNN interpretability on sketch classification

    VIEW 16 EXCERPTS
    CITES BACKGROUND & METHODS
    HIGHLY INFLUENCED

    On the Texture Bias for Few-Shot CNN Segmentation

    VIEW 1 EXCERPT
    CITES BACKGROUND

    Analyzing the Dependency of ConvNets on Spatial Information

    VIEW 4 EXCERPTS
    CITES BACKGROUND
    HIGHLY INFLUENCED

    Learning From Brains How to Regularize Machines

    VIEW 2 EXCERPTS
    CITES BACKGROUND

    pSConv: A Pre-defined S parse Kernel Based Convolution for Deep CNNs

    VIEW 1 EXCERPT
    CITES METHODS

    FILTER CITATIONS BY YEAR

    2019
    2020

    CITATION STATISTICS

    • 14 Highly Influenced Citations

    • Averaged 66 Citations per year from 2019 through 2020

    References

    Publications referenced by this paper.
    SHOWING 1-10 OF 34 REFERENCES

    Deep FisherNet for Object Classification

    VIEW 1 EXCERPT

    NetVLAD: CNN Architecture for Weakly Supervised Place Recognition

    VIEW 1 EXCERPT

    Learning Deep Features for Discriminative Localization

    VIEW 1 EXCERPT

    Deep Residual Learning for Image Recognition

    VIEW 3 EXCERPTS
    HIGHLY INFLUENTIAL

    Exploiting local features from deep networks for image retrieval

    VIEW 2 EXCERPTS