What is the best multi-stage architecture for object recognition?
Open accessThe DOI matches its Crossref record; 56 of the 57 references it cites were found in a registry or an index.
Abstract
In many recent object recognition systems, feature extraction stages are generally composed of a filter bank, a non-linear transformation, and some sort of feature pooling layer. Most systems use only one stage of feature extraction in which the filters are hard-wired, or two stages where the filters in one or both stages are learned in supervised or unsupervised mode. This paper addresses three questions: 1. How does the non-linearities that follow the filter banks influence the recognition accuracy? 2. does learning the filter banks in an unsupervised or supervised manner improve the performance over random filters or hardwired filters? 3. Is there any advantage to using an architecture with two stages of feature extraction, rather than one? We show that using non-linearities that include rectification and local contrast normalization is the single most important ingredient for good accuracy on object recognition benchmarks. We show that two stages of feature extraction yield better accuracy than one. Most surprisingly, we show that a two-stage system with random filters can yield almost 63% recognition rate on Caltech-101, provided that the proper non-linearities and pooling layers are used. Finally, we show that with supervised refinement, the system achieves state-of-the-art performance on NORB dataset (5.6%) and unsupervised pre-training followed by supervised refinement produces good accuracy on Caltech-101 (> 65%), and the lowest known error rate on the undistorted, unprocessed MNIST dataset (0.53%).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Citations
Cites
Gradient-based learning applied to document recognition
Lecun, Y., Bottou, L., Bengio, Y., & Haffner, P. (1998) · Proceedings of the IEEE · doi:10.1109/5.726791 · Cited by 58,805 · cited 6×
“Results are presented on the well-known Caltech-101 dataset [7], on the NORB object dataset [15], and on the MNIST dataset of handwritten digits [14].”
BackgroundMethodInfluentialSelf-citationDistinctive Image Features from Scale-Invariant Keypoints
Lowe, D. G. (2004) · International Journal of Computer Vision · doi:10.1023/b:visi.0000029664.99615.94 · Cited by 55,589 · cited 10×
“SIFT uses a rectification followed by winner-take-all operation over orientation , which is an extreme form of normalization.”
MethodInfluentialHistograms of Oriented Gradients for Human Detection
Dalal, N., & Triggs, B. (2005) · 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) · doi:10.1109/cvpr.2005.177 · Cited by 33,584 · cited 3×
“Particular embodiments of the single-stage systems use SIFT features [19, 13], HoG [6], Geometric Blur [5], and models inspired by the architecture of the mammalian primary visual cortex [24], to mention a few.”
MethodReducing the dimensionality of data with neural networks
Hinton, G. E., & Salakhutdinov, R. R. (2006) · Science · doi:10.1126/science.1127647 · Cited by 20,928
Harrison's Principles of Internal Medicine.
(1988) · Annals of Internal Medicine · doi:10.7326/0003-4819-108-3-499_2 · Cited by 12,514
Beyond Bags of Features: Spatial Pyramid Matching for Recognizing Natural Scene Categories
Lazebnik, S., Schmid, C., & Ponce, J. (2006) · 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 2 (CVPR'06) · doi:10.1109/cvpr.2006.68 · Cited by 7,916 · cited 12×
“The performance of our two-stage system is similar to that of the best single-stage systems based on SIFT and PMK-SVM, perhaps because the PM Kernel is conceptually similar to our feature extractions stage.”
BackgroundMethodInfluentialGreedy layer-wise training of deep networks
Bengio, Y., Lamblin, P., Popovici, D., & Larochelle, H. (2007) · Advances in Neural Information Processing Systems 19 · doi:10.7551/mitpress/7503.003.0024 · Cited by 4,841
Sparse coding with an overcomplete basis set: A strategy employed by V1?
Olshausen, B. A., & Field, D. J. (1997) · Vision Research · doi:10.1016/s0042-6989(97)00169-7 · Cited by 3,642
Linear spatial pyramid matching using sparse coding for image classification
Yang, J., Yu, K., Gong, Y., & Huang, T. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206757 · Cited by 2,828
Fei-Fei, L., Fergus, R., & Perona, P. (2005) · 2004 Conference on Computer Vision and Pattern Recognition Workshop · doi:10.1109/cvpr.2004.383 · Cited by 2,502
Cited by
Items 1,001–1,007 of 1,007
Sparse Quantization for Patch Description
Boix, X., Gygli, M., Roig, G., & Gool, L. V. (2013) · IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings · doi:10.1109/cvpr.2013.366 · Cited by 18
“Little is understood about the common principles underlying the different patch descriptors.”
BackgroundMethodInfluentialAn Instruction Set Architecture for Machine Learning
Chen, Y., Lan, H., Du, Z., Liu, S., Tao, J., Han, D., Luo, T., Guo, Q., Li, L., Xie, Y., et al. (2018) · ACM Transactions on Computer Systems · doi:10.1145/3331469 · Cited by 14
“To illustrate the usage of our proposed instruction sets, we implement three simple yet representative components of NNs, a MLP feedforward layer [68], a pooling layer [30], and a Boltzmann Machines (BM) layer [56], using Cambricon instructions.”
BackgroundMethodMacro, Finance, and Macro Finance: Solving Nonlinear Models in Continuous Time with Machine Learning
Duarte, V. (2017) · SSRN Electronic Journal · doi:10.2139/ssrn.3012602 · Cited by 9
(2014) · Cited by 33,419
(2010) · International Conference on Machine Learning · Cited by 13,995
(2011) · Cited by 6,993
Qualitatively characterizing neural network optimization problems
Goodfellow, I. J., Vinyals, O., & Saxe, A. M. (2014) · arXiv · doi:10.48550/arxiv.1412.6544 · Cited by 341
“Maxout tends to overfit on MNIST if used without dropout, so we used ReLUs for these experiments.”
MethodInfluential
Metadata
- Authors
- Jarrett, K., Kavukcuoglu, K., Ranzato, M. A., & LeCun, Y.
- Affiliations
- Courant Institute of Mathematical Sciences (US)
- ORCID
- LeCun, Yann
- Venue
- 2009 IEEE 12th International Conference on Computer Vision
- Publisher
- IEEE
- Published
- 2009-09
- Pages
- 2146-2153
- OpenAlex
- W2546302380
- Semantic Scholar
- 1f88427d7aa8225e47f946ac41a0667d7b69ac52
- Cited by
- 2,153
- References
- 57
- Citations by year
- 2026: 20 · 2025: 48 · 2024: 82 · 2023: 114 · 2022: 124 · 2021: 194 · 2020: 207 · 2019: 205 · 2018: 211 · 2017: 210 · 2016: 198 · 2015: 162 · 2014: 137 · 2013: 86 · 2012: 69
- FWCI
- 42.18
- Type
- conference paper
- Language
- en
- Open access
- green
- hal.science
- Licence
- cc-by-nc
- Locations
- IEEE/CVF International Conference on Computer Vision (ICCV) (publishedVersion)
- HAL (submittedVersion, cc-by-nc)
- Topics
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Image and Video Retrieval Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Image Retrieval and Classification Techniques
- Keywords
- object recognition, filter banks, rectification, feature pooling, unsupervised pre-training, Caltech-101, MNIST, image classification
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.