High-dimensional signature compression for large-scale image classification
The DOI matches its Crossref record; 33 of the 34 references it cites were found in a registry or an index.
Abstract
We address image classification on a large-scale, i.e. when a large number of images and classes are involved. First, we study classification accuracy as a function of the image signature dimensionality and the training set size. We show experimentally that the larger the training set, the higher the impact of the dimensionality on the accuracy. In other words, high-dimensional signatures are important to obtain state-of-the-art results on large datasets. Second, we tackle the problem of data compression on very large signatures (on the order of 105dimensions) using two lossy compression strategies: a dimensionality reduction technique known as the hash kernel and an encoding technique based on product quantizers. We explain how the gain in storage can be traded against a loss in accuracy and/or an increase in CPU cost. We report results on two large databases - ImageNet and a dataset of lM Flickr images - showing that we can reduce the storage of our signatures by a factor 64 to 128 with little loss in accuracy. Integrating the decompression in the classifier learning yields an efficient and scalable training algorithm. On ILSVRC2010 we report a 74.3% accuracy at top-5, which corresponds to a 2.5% absolute improvement with respect to the state-of-the-art. On a subset of 10K classes of ImageNet we report a top-1 accuracy of 16.7%, a relative improvement of 160% with respect to the state-of-the-art.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“since then the best published results are 45.7% and 25.7% with an approach that averages the predictions of two classifiers trained on Fisher Vectors (FVs) computed from two types of densely sampled features.29”
From ImageNet classification with deep convolutional neural networks (2017) · cites this paper as [29]
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333 · cited 1×
“org) consists of more than 12M images of 17K concepts [7] and Flickr contains thousands of groups (www.”
BackgroundBeyond 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 1×
“[15], one can split an image into several regions, compute one FV per region and concatenate the per-region FVs.”
BackgroundLocality-constrained Linear Coding for image classification
Wang, J., Yang, J., Yu, K., Lv, F., Huang, T., & Gong, Y. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5540018 · Cited by 3,100 · cited 2×
“Maji and Berg [18] and Wang et al. [31] then proposed efficient algorithms to learn IKSVMs.”
ResultProduct Quantization for Nearest Neighbor Search
Jégou, H., Douze, M., & Schmid, C. (2011) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2010.57 · Cited by 2,956 · cited 3×
“[13] proposed product quantizers (PQ) for NN-search of SIFT vectors.”
BackgroundMethodLinear 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 · cited 1×
“Yang et al. [36], Wang et al. [32] and Boureau et al. [4] showed that replacing the average pooling stage in the BOV computation by a max-pooling yielded excellent results.”
ResultWeiss, Y., Torralba, A. B., & Fergus, R. (2008) · neural information processing systems · Cited by 2,767
Aggregating local descriptors into a compact image representation
Jegou, H., Douze, M., Schmid, C., & Perez, P. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5540039 · Cited by 2,692 · cited 3×
“For instance, [23] requires dense projections and [34, 5, 14] require a global PCA.”
BackgroundMethodImproving the Fisher Kernel for Large-Scale Image Classification
Perronnin, F., Sánchez, J., & Mensink, T. (2010) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-15561-1_11 · Cited by 2,588 · cited 15×
“As shown in [22], square-rooting and L2-normalizing the FV can greatly enhance the classification accuracy.”
BackgroundMethodInfluentialSelf-citationZhang, J., Marszałek, M., Lazebnik, S., & Schmid, C. (2007) · International Journal of Computer Vision · doi:10.1007/s11263-006-9794-4 · Cited by 1,923 · cited 1×
“The winners of the PASCAL VOC 2007 [8] and 2008 [9] competitions used a similar paradigm: many types of low-level local features are extracted (referred to as “channels”), one bag-of-visual-words (BOV) histogram is computed for each channel and non-linear kernel classifiers such as SVMs are used to perform classificat…”
MethodEvaluating Color Descriptors for Object and Scene Recognition
van de Sande, K. E. A., Gevers, T., & Snoek, C. G. M. (2010) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2009.154 · Cited by 1,783 · cited 1×
“The winners of the PASCAL VOC 2007 [8] and 2008 [9] competitions used a similar paradigm: many types of low-level local features are extracted (referred to as “channels”), one bag-of-visual-words (BOV) histogram is computed for each channel and non-linear kernel classifiers such as SVMs are used to perform classificat…”
MethodGray, R. M., & Neuhoff, D. L. (1998) · IEEE Transactions on Information Theory · doi:10.1109/18.720541 · Cited by 1,642
Fisher Kernels on Visual Vocabularies for Image Categorization
Perronnin, F., & Dance, C. (2007) · 2007 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2007.383266 · Cited by 1,641 · cited 2×
“Following [20, 22] we discard the partial derivatives with respect to the mixture weights as they carry little discriminative information.”
BackgroundMethodSelf-citationDatabase-friendly random projections: Johnson-Lindenstrauss with binary coins
Achlioptas, D. (2003) · Journal of Computer and System Sciences · doi:10.1016/s0022-0000(03)00025-4 · Cited by 1,392
Learning mid-level features for recognition
Boureau, Y.-L., Bach, F., LeCun, Y., & Ponce, J. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5539963 · Cited by 1,052 · cited 1×
“[4] showed that replacing the average pooling stage in the BOV computation by a max-pooling yielded excellent results.”
ResultClassification using intersection kernel support vector machines is efficient
Maji, S., Berg, A. C., & Malik, J. (2008) · 2008 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2008.4587630 · Cited by 940 · cited 1×
“In [19], Maji et al. showed that the runtime cost of an intersection kernel (IK) SVM could be made independent of the number of support vectors.”
BackgroundFeature hashing for large scale multitask learning
Weinberger, K., Dasgupta, A., Langford, J., Smola, A., & Attenberg, J. (2009) · Proceedings of the 26th Annual International Conference on Machine Learning · doi:10.1145/1553374.1553516 · Cited by 906 · cited 6×
“They have shown excellent performance in the compression of text features [26, 33] and faces [25].”
BackgroundMethodInfluentialSmall codes and large image databases for recognition
Torralba, A., Fergus, R., & Weiss, Y. (2008) · 2008 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2008.4587633 · Cited by 727 · cited 1×
“showed that small codes enable efficient image retrieval and k-NN classification on a large scale [27].”
BackgroundLocality-sensitive binary codes from shift-invariant kernels
Raginsky, M., & Lazebnik, S. (2009) · Cited by 673
Image Classification Using Super-Vector Coding of Local Image Descriptors
Zhou, X., Yu, K., Zhang, T., & Huang, T. S. (2010) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-15555-0_11 · Cited by 499 · cited 2×
“PASCAL VOC 2009 winners mention in [10] (slide 27) that their GMM-based [39] and sparse-coding-based [37] representations have respectively 655K and 2M dimensions.”
BackgroundHuman detection using partial least squares analysis
Schwartz, W. R., Kembhavi, A., Harwood, D., & Davis, L. S. (2009) · 2009 IEEE 12th International Conference on Computer Vision · doi:10.1109/iccv.2009.5459205 · Cited by 493 · cited 2×
“2 in [24]) but should deteriorate performance for a large number of classes.”
BackgroundWhat Does Classifying More Than 10,000 Image Categories Tell Us?
Deng, J., Berg, A. C., Li, K., & Fei-Fei, L. (2010) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-15555-0_6 · Cited by 486 · cited 2×
“We finally replicated the very large-scale experiments of Deng et al. [6]: we ran our system on the same 10,184 categories from the Fall 2009 release of ImageNet including both internal and leaf nodes with more than 200 images .”
Efficient Object Category Recognition Using Classemes
Torresani, L., Szummer, M., & Fitzgibbon, A. (2010) · Lecture Notes in Computer Science · doi:10.1007/978-3-642-15549-9_56 · Cited by 409 · cited 2×
“Other works have proposed to perform a semantic dimensionality reduction [31, 28].”
BackgroundEfficient additive kernels via explicit feature maps
Vedaldi, A., & Zisserman, A. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5539949 · Cited by 257
Landmark classification in large-scale image collections
Li, Y., Crandall, D. J., & Huttenlocher, D. P. (2009) · 2009 IEEE 12th International Conference on Computer Vision · doi:10.1109/iccv.2009.5459432 · Cited by 244 · cited 2×
“[16] study landmark classification on a collection of 500 landmarks and 2 million images.”
BackgroundJegou, H., Douze, M., & Schmid, C. (2009) · 2009 IEEE 12th International Conference on Computer Vision · doi:10.1109/iccv.2009.5459419 · Cited by 159 · cited 3×
“The aggregation process may be unbalanced [26, 33] or balanced [12], i.”
BackgroundYan, S., Zhou, X., Liu, M., Hasegawa-Johnson, M., & Huang, T. S. (2008) · 2008 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2008.4587405 · Cited by 154 · cited 1×
“This includes modeling an image by a probability distribution [17, 35] or using the Fisher kernel framework [20].”
MethodMax-margin additive classifiers for detection
Maji, S., & Berg, A. C. (2009) · 2009 IEEE 12th International Conference on Computer Vision · doi:10.1109/iccv.2009.5459203 · Cited by 153 · cited 1×
“Maji and Berg [18] and Wang et al. [31] then proposed efficient algorithms to learn IKSVMs.”
BackgroundLarge-scale image categorization with explicit data embedding
Perronnin, F., Sanchez, J., & Liu, Y. (2010) · 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2010.5539914 · Cited by 152 · cited 1×
“[21] subsequently generalized this principle to any additive classifier.”
BackgroundSelf-citationTransform coding for fast approximate nearest neighbor search in high dimensions
Brandt, J. (2010) · 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.2010.5539852 · Cited by 116 · cited 2×
“For instance, [23] requires dense projections and [34, 5, 14] require a global PCA.”
BackgroundEfficient Highly Over-Complete Sparse Coding Using a Mixture Model
Yang, J., Yu, K., & Huang, T. S. (2010) · Lecture notes in computer science · doi:10.1007/978-3-642-15555-0_9 · Cited by 105 · cited 1×
“PASCAL VOC 2009 winners mention in [10] (slide 27) that their GMM-based [39] and sparse-coding-based [37] representations have respectively 655K and 2M dimensions.”
BackgroundLearning image similarity from Flickr groups using Stochastic Intersection Kernel MAchines
Wang, G., Hoiem, D., & Forsyth, D. (2009) · 2009 IEEE 12th International Conference on Computer Vision · doi:10.1109/iccv.2009.5459167 · Cited by 84 · cited 4×
“com/groups) – some of which with hundreds of thousands of pictures – which can be readily used to learn object classifiers [31, 22].”
BackgroundMethodInfluentialRapid face recognition using hashing
Shi, Q., Li, H., & Shen, C. (2010) · 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.2010.5540001 · Cited by 59 · cited 1×
“They have shown excellent performance in the compression of text features [26, 33] and faces [25].”
BackgroundA similarity measure between unordered vector sets with application to image categorization
Liu, Y., & Perronnin, F. (2008) · 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.2008.4587600 · Cited by 24 · cited 1×
“This includes modeling an image by a probability distribution [17, 35] or using the Fisher kernel framework [20].”
MethodSelf-citation
Research works printed but not linked: 6
- shi, 2009, Hash kernels, AISTATS
- bottou, 0, SGD
- berg, 2010, ILSVRC
- gong, 2009, Image classification using Gaussian mixture and local coordinate coding, PASCAL VOC 2009 Workshop
- everingham, 0, The PASCAL VOC 2008 Results
- everingham, 2007, The PASCAL VOC
Cited by
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
“7% with an approach that averages the predictions of two classifiers trained on Fisher Vectors (FVs) computed from two types of densely sampled features.(29) We also entered our model in the ILSVRC-2012 competition and report our results in Table 2.”
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380
“Im a g eN et L a rg e S ca le V isu a l R eco g n itio n C h a llen g e 1 7 ILSVRC 2010 Codename CLS LOC Insitutions Contributors and references Hminmax 54.4 Massachusetts Institute of Technology Jim Mutch, Sharat Chikkerur, Hristo Paskov, Ruslan Salakhutdinov, Stan Bileschi, Hueihan Jhuang IBM 70.1 IBM research†, Geo…”
BackgroundMethodInfluential(2014) · Cited by 33,419
Improving neural networks by preventing co-adaptation of feature detectors
Hinton, G. E., Srivastava, N., Krizhevsky, A., Sutskever, I., & Salakhutdinov, R. R. (2012) · arXiv · doi:10.48550/arxiv.1207.0580 · Cited by 6,382
“7% [16].”
Deep Convolutional Neural Networks for Image Classification: A Comprehensive Review
Rawat, W., & Wang, Z. (2017) · Neural Computation · doi:10.1162/neco_a_00990 · Cited by 3,503
“Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/neco_a_00990 by guest on 04 June 2021 Lin et al., 2011; Sánchez & Perronnin, 2011), as well as an approximate correlation between classification performance and network depth (see section 5.5.4).”
BackgroundInfluential
Metadata
- Authors
- Sanchez, J., & Perronnin, F.
- Affiliations
- Xerox (France) (FR)
- Venue
- CVPR 2011
- Publisher
- IEEE
- Published
- 2011-06
- Pages
- 1665-1672
- ISSN
- 1063-6919 (ISSN-L)
- OpenAlex
- W2053229256
- Semantic Scholar
- eefcc7bcc05436dac9881acb4ff4e4a0b730e175
- Cited by
- 272
- References
- 34
- Citations by year
- 2026: 3 · 2025: 5 · 2024: 7 · 2023: 6 · 2022: 11 · 2021: 19 · 2020: 10 · 2019: 23 · 2018: 17 · 2017: 23 · 2016: 21 · 2015: 46 · 2014: 30 · 2013: 30 · 2012: 18
- FWCI
- 19.28
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings - IEEE Computer Society Conference on Computer Vision and Pattern Recognition/Proceedings (publishedVersion)
- 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 › Image Retrieval and Classification Techniques
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- image classification, large-scale image classification, signature compression, lossy compression, dimensionality reduction, training data size, storage efficiency, scalable training, ImageNet, Flickr images
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.