Adaptive deconvolutional networks for mid and high level feature learning

By Matthew D. Zeiler, Graham W. Taylor, Rob Fergus. 2011

1,302 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

ImageNet Large Scale Visual Recognition Challenge (2014, methodology): “The network architectures were chosen using the visualization technique of (Zeiler and Fergus, 2013), and they were trained on the GPU following (Zeiler et al., 2011) using the dropout technique (Krizhevsky et al., 2012).”
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation (2013, neutral): “The final two layers are summarized below.”
Visualizing and Understanding Convolutional Networks (2013, methodology): “We perform this mapping with a Deconvolutional Network (deconvnet) (Zeiler et al., 2011).”
Unified Perceptual Parsing for Scene Understanding (2018, background): “Noh et al. [12] propose a framework which applies deconvolution [13] to up-sample low resolution feature maps.”
Learning to generate chairs with convolutional neural networks (2014, neutral): “This is similar to the “deconvolutional” layers used in previous work [32, 11, 33].”
Learning Energy-Based Spatial-Temporal Generative ConvNets for Dynamic Patterns (2019, methodology): “The back-propagation process defines a top-down deconvolution process [50], where the filters at multiple layers become the basis functions at those layers, and the activation variables at different layers in dðI; uÞ become the coefficients of the basis functions in the topdown deconvolution.”
A Theory of Generative ConvNet (2016, background): “3; Xie et al., 2014). The generative ConvNet may help solve this problem. The representational structure in the generative ConvNet is similar to but subtly different from the deconvolution network of Zeiler et al. (2011). The top-down process of the generative ConvNet is controlled by multiple layer”
Generating images with recurrent adversarial networks (2016, background): “, 2015) showed excellent samples of realistic images using fully convolutional neural network as the discriminative model and fully decovolutional neural network (Zeiler et al., 2011) as the generative model.”
Unsupervised and Semi-supervised Learning with Categorical Generative Adversarial Networks (2015, methodology): “Finally, we performed experiments with convolutional discriminator networks and de-convolutional (Zeiler et al., 2011) generator networks (using the same up-sampling procedure from Dosovitskiy et al. (2015)) on MNIST and CIFAR-10.”
Semi-supervised Learning with Ladder Networks (2015, background): “Taken together, these results suggest that combining the generalization ability of convolutional networks and efﬁcient unsupervised learning of the full Ladder network would have resulted in even better performance but this was left for future work.”
Rethinking Atrous Convolution for Semantic Image Segmentation (2017, methodology): “For example, [53, 56] employ deconvolution [83] to learn the upsampling of low resolution feature responses.”
Learning Deconvolution Network for Semantic Segmentation (2015, methodology): “To implement the unpooling operation, we follow the similar approach proposed in [26, 27].”
Exploiting Linear Structure Within Convolutional Networks for Efficient Evaluation (2014, background): “ propagation pass U(I n; ), and dene as f ngthe indices of the hlargest values of U(I n; ) dierent from y n. Then, for a given layer s, we compute d n;l;s= r W s (U(I n; ) (i l)) ; nN;l2f ng;sS; (8) Title Suppressed Due to Excessive Length 9 where (i l) is the dirac distribution centered at l. In ot”
Explainable Artificial Intelligence (XAI): Concepts, Taxonomies, Opportunities and Challenges toward Responsible AI (2019, positive): “One of the seminal works in the ﬁrst category was [291] .”
Towards Better Analysis of Deep Convolutional Neural Networks (2016, methodology): “For example, Zeiler et al. [58] utilized a multi-layered Deconvolutional Network [59] to project the activations onto the input pixel space.”
Deep Learning: Methods and Applications (2014, methodology): “[437], Zeiler and Fergus [435, 436], and Zeiler ([434]).”