Evaluation of Pooling Operations in Convolutional Architectures for Object Recognition

By Dominik Scherer, Andreas C. Müller, Sven Behnke. 2010

1,636 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

Making Convolutional Networks Shift-Invariant Again (2019, methodology): “However, ample empirical evidence suggests max-pooling provides stronger task performance (Scherer et al., 2010), leading to its widespread adoption.”
On Buggy Resizing Libraries and Surprising Subtleties in FID Calculation (2021, background): “While it was replaced by operations such as max-pooling, based on empirical performance [65], recent works have demonstrated that antialiasing can be compatible and improve performance in convolutional networks [79,85], transformers [57], NeRFs [4], and GANs [36].”
Multi-column deep neural networks for image classification (2012, methodology): “[8]; 5.60% - [32] No preprocessing is used for this dataset.”
Deep learning in neural networks: An overview (2014, methodology): “Advantages of doing this were pointed out subsequently (Scherer et al., 2010).”
Lifelong learning of human actions with deep neural network self-organization (2017, background): “the MAX operator showing faster convergence and improved generalization in deep neural network architectures (Scherer et al., 2010).”
High-Performance Neural Networks for Visual Object Classification (2011, methodology): “The CNN architecture considered in this study diﬀers from the one of Simard et al. (2003) in the sense that after each CNN-layer an optional max-pooling layer (Scherer et al. , 2010) can be used.”
Generalizing Pooling Functions in Convolutional Neural Networks: Mixed, Gated, and Tree (2015, background): “choosing from more than just one type of pooling operation can benefit performance [31].”
Deep Networks with Internal Selective Attention through Feedback Connections (2014, neutral): “Deep convolutional neural networks (CNNs) [1] with max-pooling layers [2] trained by backprop [3–5] on GPUs [6] have become the state-of-the-art in object recognition [7–10], segmentation/detection [11, 12], and scene parsing [13–15] (for an extensive review see [16]).”
On Learning to Think: Algorithmic Information Theory for Novel Combinations of Reinforcement Learning Controllers and Recurrent Neural World Models (2015, neutral): “All weights are trained by backpropagation [140, 199, 220, 245].”
One Big Net For Everything (2018, background): “If there are large 2-dimensional inputs such as video images, the LSTM may have a front-end [95] in form of a convolutional neural net (CNN) [11, 33, 97, 4, 45, 52, 8, 76] implemented on fast GPUs [8, 76].”
Oriented Response Networks (2017, background): “Such capability is endowed with the inherent properties of convolutional operations, redundant convolutional filters, and hierarchical spatial pooling [35, 20].”
Deep 3D Convolutional Encoder Networks With Shortcuts for Multiscale Feature Integration Applied to Multiple Sclerosis Lesion Segmentation (2016, methodology): “A convolutional layer is followed by an average pooling layer [33] that halves the number of units in each dimension by calculating the average of each block of 2 2 2 units per channel.”
Multi-Classification of Brain Tumor Images Using Deep Neural Network (2019, methodology): “The pooling layer is used to reduce numbers of parameters and consequently computations in the network [34], [35].”
On Fast Deep Nets for AGI Vision (2011, background): “We sometimes (but not always) proﬁt from sparse network connectivity and techniques such as weight sharing & convolution [12, 1, 25], max-pooling [17], and contrast enhancement [6] like the one automatically generated by unsupervised Predictability Minimization [18, 22, 24].”
Deep learning for remote sensing image classification: A survey (2018, background): “Detailed theoretical analysis of these is beyond the scope of this paper, but can be found in Scherer, Muller, and Behnke (2010).”
Deep learning in vision-based static hand gesture recognition (2017, background): “in the networks as obtains in some other related works [30, 32].”
An CNN-LSTM Attention Approach to Understanding User Query Intent from Online Health Communities (2017, methodology): “2) Max-Pooling layer: We apply a max-over-time pooling operation [19] with the feature maps built by the CNN filters.”