Beyond spatial pyramids: Receptive field learning for pooled image features

By Yangqing Jia, Chang Huang, Trevor Darrell. 2012

304 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

In Representation & Self-Supervised Learning

Citations in the corpus

Spectral Networks and Locally Connected Networks on Graphs (2013, background): “In [3, 12] the authors estimate similarities between features to construct locally connected networks.”
Deformable Convolutional Networks (2017, background): “The work in [25] learns a sparse subset of pooling regions from a large over-complete set.”
Striving for Simplicity: The All Convolutional Net (2014, methodology): “We emphasize that that this replacement can also be seen as learning the pooling operation rather than fixing it; which has previously been considered using different parameterizations in the literature 2 (LeCun et al., 1998; Gülçehre et al., 2014; Jia et al., 2012).”
Fast ConvNets Using Group-Wise Brain Damage (2015, background): “to discover the optimal structure of a graphical model [22] or the optimal receptive fields in the two layered image classifier [25].”
Stochastic Pooling for Regularization of Deep Convolutional Neural Networks (2013, background): “Train Error % Test Error % Receptive Field Learning [4] – 45.”
Deep roto-translation scattering for object classification (2014, methodology): “Dict 77.3 RFL [8] Unsup.”
Joint Unsupervised Learning of Deep Representations and Image Clusters (2016, methodology): “This performance also beats several other methods listed in [4, 15, 25, 34].”
Large-Scale Feature Learning With Spike-and-Slab Sparse Coding (2012, methodology): “8 ± 1% (Jia and Huang, 2011), achieved using a learned pooling structure on top of “triangle code” features from a dictionary learned using k-means.”
On the Expressive Power of Overlapping Architectures of Deep Learning (2017, background): “Other works studying the receptive fields of neural networks have mainly focused on how to learn them from the data (Coates and Ng, 2011; Jia et al., 2012).”
Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation (2014, background): “pyramids[22]), or in a fixed but learned configuration[19], both independent of the image content observed.”
Learnable Pooling Regions for Image Classification (2013, methodology): “For instance Jia and Huang [2011] optimizes binary pooling strategies that are given by the superposition of rectangular basis functions, and Feng et al. [2011] ﬁnds pooling regions by applying a linear discriminant analysis for individual pooling strategies and training a classiﬁer afterwards.”
Deep Regionlets for Object Detection (2017, methodology): “If we only enforce the region selection network to learn the shift, our regionlet learning mechanism would degenerate to similar deformable RoI pooling as in [6, 32] Spatial-based RoI pooling [24, 22, 17].”
Learning Smooth Pooling Regions for Visual Recognition (2013, methodology): “We arrive at a test accuracy of 80.35% which is an additional improvement of 0.75% and 0.18% over our best results (batch-based approximation) and Jia and Huang [10] respectively.”
Hierarchical Convolutional Deep Learning in Computer Vision (2013, background): “Train Error % Test Error % Receptive Field Learning [57] – 45.”
Generalizing Pooling Functions in CNNs: Mixed, Gated, and Tree (2018, methodology): “Learning pooling functions is analogous to receptive field learning [17], [18], [19], [20].”
A survey of deep neural network architectures and their applications (2017, background): “In [78], it is shown that better po oling performance can be achieved by learning receptive”