Learning mid-level features for recognition

By Y-Lan Boureau, Francis R. Bach, Yann LeCun, Jean Ponce. 2010

1,157 citations recorded by the source corpus.

Subfields: Aerospace Engineering, Computer Vision and Pattern Recognition, Representation & Self-Supervised Learning

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Aerospace Engineering

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Character-level Convolutional Networks for Text Classification (2015, methodology): “It is the 1-D version of the max-pooling module used in computer vision [2].”
Text Understanding from Scratch (2015, methodology): “The analysis by (Boureau et al., 2010b) might shed some light on this.”
A Sensitivity Analysis of (and Practitioners’ Guide to) Convolutional Neural Networks for Sentence Classification (2015, background): “Next, we considered taking an average, rather than the max, over regions (Boureau et al., 2010a).”
Invariant Scattering Convolution Networks (2012, background): “…exists ǫ > 0 such that for all ω ∈ R 2 then applying the Plancherel formula proves that W J x = { x ⋆ φ J , x ⋆ ψ λ } λ ∈ Λ J satisﬁes with k k k k P k k In the following we suppose that ǫ < 1 and hence that the wavelet transform is a contracting and invertible operator, with a stable inverse.”
Learning Fine-Grained Image Similarity with Deep Ranking (2014, methodology): “One way to build image similarity models is to first extract features like Gabor filters, SIFT [17] and HOG [4], and then learn the image similarity models on top of these features [2, 3, 22].”
Emergence of Complex-Like Cells in a Temporal Product Network with Local Receptive Fields (2010, background): “These results are not state of the art which is currently for systems of this class is 75% Boureau et al. (2010) (the system extracts sift features, then learns sparse dictionaries, pools and makes histograms).”
A guide to convolution arithmetic for deep learning (2016, background): “Other kinds of pooling exist, e.g., mean or average pooling, which all share the same idea of aggregating the input locally by applying a non-linearity to the content of some patches (Boureau et al. , 2010a,b, 2011; Saxe et al. , 2011).”
Mining Mid-level Features for Image Classification (2014, methodology): “[6,14,22]) starts from the same description that proved its value for local features, building on robust gradient-based representations such as HoG [13].”
Author manuscript, published in "International Journal of Computer Vision (2013)" International Journal of Computer Vision manuscript No. (will be inserted by the editor) Image Classification with the Fisher Vector: Theory and Practice (2013, positive): “Boureau et al (2010), Wang et al (2010), Yang et al (2009b) showed that replacing the average pooling stage in the BoV computation by a max-pooling yielded excellent results.”
Deep convolutional neural networks for brain image analysis on magnetic resonance imaging: a review (2017, background): “Max pooling is the most common due to its empirical performance [64, 65].”
Image Reconstruction from Bag-of-Visual-Words (2014, methodology): “Instead of assigning each local descriptor to one visual word (hard assignment), some method [33], [34], [35], [36], [37], [38] represents each descriptor by a vector, the i-th dimension of which is the weight for the i-th visual word.”
Latent Dirichlet Allocation Models for Image Classification (2013, methodology): “It is the foundation of a number of methods for object recognition and image classification [14], [27], [6].”
Weakly Supervised Object Localization with Latent Category Learning (2014, methodology): “Many recent studies use the Bag-of-Words (BoW) feature for its mid-level object representation [3, 9, 23, 28, 29], and some”
Text Detection and Character Recognition in Scene Images with Unsupervised Feature Learning (2011, background): “A common way to do this is with spatial pooling [26] where we combine the responses of a feature at multiple locations into a single feature.”
Recurrent convolutional neural network for object recognition (2015, background): “Note that supervised learning techniques can be incorporated into the unsupervised learning framework of sparse coding models [3].”
Ask the locals: Multi-way local pooling for image recognition (2011, methodology): “It has recently gained popularity due to its better performance when paired with sparse coding and simple linear classiﬁers [4, 38, 41], and its statistical properties which make it well suited to sparse representations [5].”
Super Normal Vector for Activity Recognition Using Depth Sequences (2014, background): “For example, the spatial neighborhoods of low-level features are jointly encoded in deep learning [8] and macrofeatures [1].”