Elamur Research
☰
home / papers / 2014 / Very Deep Convolutional Networks for Large-Scale Image Recognition

Very Deep Convolutional Networks for Large-Scale Image Recognition

Written for the Research library (claude-opus-5-5) from the paper text, in our own words: the license "arXiv — non-exclusive licence to distribute" does not allow redistributing the text itself. Original — arxiv.org/pdf/1409.1556

Summary

The paper looks at one question in convolutional network design: how much does depth alone help on large-scale image classification? Simonyan and Zisserman keep the rest of the design fixed. Every convolution uses a 3×3 filter with stride 1, there are five max-pooling stages, and three fully connected layers sit on top. They then train five configurations (A–E) with 11 to 19 weight layers on ImageNet (ILSVRC). Accuracy goes up steadily as depth increases, and the 16- and 19-layer networks clearly beat earlier, shallower designs. A stack of small filters covers the same receptive field as one large filter but uses fewer parameters and adds more non-linearities. This work was the basis of the team's ILSVRC-2014 entry, which won the localisation track and came second in classification. The authors also took features from the trained networks and fed them to simple linear classifiers on other benchmarks, where they set new state-of-the-art results. They released the two best models publicly. These models, now known as VGG-16 and VGG-19, became standard backbones for detection, segmentation and captioning.

Key points

  • Configurations range from net A (8 conv + 3 FC layers) to net E (16 conv + 3 FC). Channel width grows from 64 to 512, and the parameter count barely changes, from 133M to 144M.
  • Three stacked 3×3 layers cover a 7×7 receptive field with 27C² weights. A single 7×7 layer would need 49C². Local Response Normalisation (LRN) gave no accuracy gain.
  • Training used SGD with momentum (batch 256), weight decay and dropout, and stopped after about 370K iterations (74 epochs). The deeper nets were initialised from the layers of net A.
  • ILSVRC-2014: 1st place in localisation and 2nd in classification. The paper says a single VGG network beats a single GoogLeNet network on classification.
  • Transfer results: more than 6% above the previous best on PASCAL VOC among ILSVRC-pretrained representations, 8.6% above the prior best on Caltech-256, and 84.0 mAP on VOC-2012 action classification.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

Fields: Other

Cited by

1394 papers of our library cite it (of the 140673 whose bibliographies we read); 37 of them have a page here, 1357 not yet.

Cite

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.1409.1556 · arXiv 1409.1556

@misc{simonyan2014very,
  title = {{Very Deep Convolutional Networks for Large-Scale Image Recognition}},
  author = {Simonyan, Karen and Zisserman, Andrew},
  year = {2014},
  doi = {10.48550/arxiv.1409.1556},
  eprint = {1409.1556},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.1409.1556}
}

References

34 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Bell, S., Upchurch, P., Snavely, N., and Bala, K. Material recognition in the wild with the materials in context database. CoRR, abs/1412.0623, 2014. — [arXiv:1412.0623](https://arxiv.org/abs/1412.0623) find
  2. Chatfield, K., Simonyan, K., Vedaldi, A., and Zisserman, A. Return of the devil in the details: Delving deep into convolutional nets. In Proc. BMVC., 2014. find
  3. Cimpoi, M., Maji, S., and Vedaldi, A. Deep convolutional filter banks for texture recognition and segmentation. CoRR, abs/1411.6836, 2014. — [arXiv:1411.6836](https://arxiv.org/abs/1411.6836) find
  4. Ciresan, D. C., Meier, U., Masci, J., Gambardella, L. M., and Schmidhuber, J. Flexible, high performance convolutional neural networks for image classification. In IJCAI, pp. 1237–1242, 2011. find
  5. Dean, J., Corrado, G., Monga, R., Chen, K., Devin, M., Mao, M., Ranzato, M., Senior, A., Tucker, P., Yang, K., Le, Q. V., and Ng, A. Y. Large scale distributed deep networks. In NIPS, pp. 1232–1240, 2012. find
  6. Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., and Fei-Fei, L. Imagenet: A large-scale hierarchical image database. In Proc. CVPR, 2009. find
  7. Donahue, J., Jia, Y., Vinyals, O., Hoffman, J., Zhang, N., Tzeng, E., and Darrell, T. Decaf: A deep convolutional activation feature for generic visual recognition. CoRR, abs/1310.1531, 2013. find
  8. Everingham, M., Eslami, S. M. A., Van Gool, L., Williams, C., Winn, J., and Zisserman, A. The Pascal visual object classes challenge: A retrospective. IJCV, 111(1):98–136, 2015. find
  9. Fei-Fei, L., Fergus, R., and Perona, P. Learning generative visual models from few training examples: An incremental bayesian approach tested on 101 object categories. In IEEE CVPR Workshop of Generative Model Based Vision, 2004. find
  10. Girshick, R. B., Donahue, J., Darrell, T., and Malik, J. Rich feature hierarchies for accurate object detection and semantic segmentation. CoRR, abs/1311.2524v5, 2014. Published in Proc. find our card
  11. Gkioxari, G., Girshick, R., and Malik, J. Actions and attributes from wholes and parts. CoRR, abs/1412.2604, 2014. — [arXiv:1412.2604](https://arxiv.org/abs/1412.2604) find
  12. Glorot, X. and Bengio, Y. Understanding the difficulty of training deep feedforward neural networks. In Proc. AISTATS, volume 9, pp. 249–256, 2010. find
  13. Goodfellow, I. J., Bulatov, Y., Ibarz, J., Arnoud, S., and Shet, V. Multi-digit number recognition from street view imagery using deep convolutional neural networks. In Proc. ICLR, 2014. find
  14. Griffin, G., Holub, A., and Perona, P. Caltech-256 object category dataset. Technical Report 7694, California Institute of Technology, 2007. find
  15. He, K., Zhang, X., Ren, S., and Sun, J. Spatial pyramid pooling in deep convolutional networks for visual recognition. CoRR, abs/1406.4729v2, 2014. — [arXiv:1406.4729](https://arxiv.org/abs/1406.4729) find
  16. Hoai, M. Regularized max pooling for image categorization. In Proc. BMVC., 2014. find
  17. Howard, A. G. Some improvements on deep convolutional neural network based image classification. In Proc. ICLR, 2014. find
  18. Jia, Y. Caffe: An open source convolutional architecture for fast feature embedding. http://caffe.berkeleyvision.org/, 2013. ↗
  19. Karpathy, A. and Fei-Fei, L. Deep visual-semantic alignments for generating image descriptions. CoRR, abs/1412.2306, 2014. find our card
  20. Kiros, R., Salakhutdinov, R., and Zemel, R. S. Unifying visual-semantic embeddings with multimodal neural language models. CoRR, abs/1411.2539, 2014. find our card
  21. Krizhevsky, A. One weird trick for parallelizing convolutional neural networks. CoRR, abs/1404.5997, 2014. find
  22. Krizhevsky, A., Sutskever, I., and Hinton, G. E. ImageNet classification with deep convolutional neural networks. In NIPS, pp. 1106–1114, 2012. find
  23. LeCun, Y., Boser, B., Denker, J. S., Henderson, D., Howard, R. E., Hubbard, W., and Jackel, L. D. Backpropagation applied to handwritten zip code recognition. Neural Computation, 1(4):541–551, 1989. find
  24. Lin, M., Chen, Q., and Yan, S. Network in network. In Proc. ICLR, 2014. find
  25. Long, J., Shelhamer, E., and Darrell, T. Fully convolutional networks for semantic segmentation. CoRR, abs/1411.4038, 2014. — [arXiv:1411.4038](https://arxiv.org/abs/1411.4038) find
  26. Oquab, M., Bottou, L., Laptev, I., and Sivic, J. Learning and Transferring Mid-Level Image Representations using Convolutional Neural Networks. In Proc. CVPR, 2014. find
  27. Perronnin, F., Sánchez, J., and Mensink, T. Improving the Fisher kernel for large-scale image classification. In Proc. ECCV, 2010. find
  28. Razavian, A., Azizpour, H., Sullivan, J., and Carlsson, S. CNN Features off-the-shelf: an Astounding Baseline for Recognition. CoRR, abs/1403.6382, 2014. — [arXiv:1403.6382](https://arxiv.org/abs/1403.6382) find
  29. Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., Berg, A. C., and Fei-Fei, L. ImageNet large scale visual recognition challenge. CoRR, abs/1409.0575, 2014. find our card
  30. Sermanet, P., Eigen, D., Zhang, X., Mathieu, M., Fergus, R., and LeCun, Y. OverFeat: Integrated Recognition, Localization and Detection using Convolutional Networks. In Proc. ICLR, 2014. find
  31. Simonyan, K. and Zisserman, A. Two-stream convolutional networks for action recognition in videos. CoRR, abs/1406.2199, 2014. Published in Proc. — [arXiv:1406.2199](https://arxiv.org/abs/1406.2199) find
  32. Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D., Vanhoucke, V., and Rabinovich, A. Going deeper with convolutions. CoRR, abs/1409.4842, 2014. — [arXiv:1409.4842](https://arxiv.org/abs/1409.4842) find
  33. Wei, Y., Xia, W., Huang, J., Ni, B., Dong, J., Zhao, Y., and Yan, S. CNN: Single-label to multi-label. CoRR, abs/1406.5726, 2014. — [arXiv:1406.5726](https://arxiv.org/abs/1406.5726) find
  34. Zeiler, M. D. and Fergus, R. Visualizing and understanding convolutional networks. CoRR, abs/1311.2901, 2013. Published in Proc. find our card