CoChat

Dynamic Convolutions: Exploiting Spatial Sparsity for Faster Inference

Open access
Verelst, Thomas, Tuytelaars, Tinne · 2020 · 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Cited by 170

The DOI matches its Crossref record; 22 of the 22 references it cites were found in a registry or an index.

Abstract

Modern convolutional neural networks apply the same operations on every pixel in an image. However, not all image regions are equally important. To address this inefficiency, we propose a method to dynamically apply convolutions conditioned on the input image. We introduce a residual block where a small gating branch learns which spatial positions should be evaluated. These discrete gating decisions are trained end-to-end using the Gumbel-Softmax trick, in combination with a sparsity criterion. Our experiments on CIFAR, ImageNet, Food-101 and MPII show that our method has better focus on the region of interest and better accuracy than existing methods, at a lower computational complexity. Moreover, we provide an efficient CUDA implementation of our dynamic convolutions using a gather-scatter approach, achieving a significant improvement in inference speed on MobileNetV2 and ShuffleNetV2. On human pose estimation, a task that is inherently spatially sparse, the processing speed is increased by 60% with no loss in accuracy.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1109/cvpr42600.2020.00239

Cites

22 of 22 references verified0 citation errors · 0 not checkable

Statistics

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 22 references

Research works printed but not linked: 33

  1. molchanov, 2016, Pruning convolutional neural networks for resource efficient inference
  2. mnih, 2014, Recurrent models of visual attention, Advances in Neural Information Processing Systems (NIPS), 2204
  3. maddison, 0, The concrete distribution: A continuous relaxation of discrete random variables, Proc of the Int Conf on Learning Representations (ICLR)
  4. lin, 2017, Runtime neural pruning, Advances in Neural Information Processing Systems (NIPS), 2181
  5. 10.3850/9783981537079_0819 · DOI · Paper page
  6. lecun, 1990, Optimal brain damage, Advances in Neural Information Processing Systems (NIPS), 598
  7. 10.1109/CVPR.2017.684 · DOI · Paper page
  8. 10.1109/CVPR.2014.471 · DOI · Paper page
  9. almahairi, 0, Dynamic capacity networks, Proceedings of the International Conference on Machine Learning (ICML), 2549
  10. 10.1162/neco.1991.3.1.79 · DOI · Paper page
  11. kim, 2015, Compression of deep convolutional neural networks for fast and low power mobile applications
  12. 10.1109/CVPR.2016.435 · DOI · Paper page
  13. krizhevsky, 2009, Learning multiple layers of features from tiny images Technical report
  14. 10.1109/CVPR.2016.521 · DOI · Paper page
  15. 10.1109/CVPR.2019.00363 · DOI · Paper page
  16. 10.1016/j.robot.2014.11.005 · DOI
  17. sharma, 2015, Action recognition using visual attention
  18. 10.1109/CVPR.2017.194 · DOI · Paper page
  19. gao, 2018, Dynamic channel pruning Feature boosting and suppression
  20. graham, 2014, Spatially-sparse convolutional neural networks
  21. graham, 2017, Sub-manifold sparse convolutional networks
  22. han, 2015, Deep compression Compressing deep neural networks with pruning trained quantization and huffman coding
  23. bengio, 2015, Conditional computation in neural networks for faster models
  24. bejnordi, 2019, Batch-shaped channel gated networks
  25. 10.1007/978-3-642-39593-2_1 · DOI · Paper page
  26. bossard, 0, Food-10 1–mining discriminative components with random forests, Proceedings of the European Conference on Computer Vision (ECCV), 446
  27. bolukbasi, 0, Adaptive neural networks for efficient inference, Proc of the International Conference on Machine Learning (ICML), 70, 527
  28. wen, 2016, Learning structured sparsity in deep neural networks, Advances in Neural Information Processing Systems (NIPS), 2074
  29. 10.1007/978-3-030-01246-5_1 · DOI · Paper page
  30. wang, 0, Skipnet: Learning dynamic routing in convolutional networks, Proceedings of the European Conference on Computer Vision (ECCV), 409
  31. veit, 2016, Residual networks behave like ensembles of relatively shallow networks, Advances in Neural Information Processing Systems (NIPS), 550
  32. sindhwani, 2015, Structured transforms for small-footprint deep learning, Advances in Neural Information Processing Systems (NIPS), 3088
  33. 10.1109/CVPR.2016.308 · DOI · Paper page

Cited by

Metadata

Authors
Verelst, T., & Tuytelaars, T.
Venue
2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
Publisher
IEEE
Published
2020-06
Pages
2317-2326
ISSN
1063-6919 (ISSN-L)
OpenAlex
W2991680559
Cited by
170
References
22
FWCI
7.54
Type
conference paper
Language
en
Open access
green

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.