The Shape Boltzmann Machine: A Strong Model of Object Shape

By S. M. Ali Eslami, Nicolas Manfred Otto Heess, Christopher K. I. Williams, John M. Winn. 2012

208 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Generative Models

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

In Generative Models

Citations in the corpus

Deep Generative Image Models using a Laplacian Pyramid of Adversarial Networks (2015, methodology): “Restricted Boltzmann machines [14, 18, 21, 23], Deep Boltzmann machines [26, 8], Denoising auto-encoders [30] all have a generative decoder that reconstructs the image from the latent representation.”
Learning to See by Moving (2015, methodology): “The coordinates of the keypoints are provided in the image coordinate system [8].”
Unsupervised Learning of Visual Representations Using Videos (2015, background): “The third class of algorithms and more related to our paper is unsupervised learning of visual representations from the pixels themselves using deep learning approaches [21, 26, 44, 39, 29, 47, 9, 33, 2, 49, 8].”
Attend, Infer, Repeat: Fast Scene Understanding with Generative Models (2016, background): “…range from highly speciﬁed (Milch et al., 2005; Mansinghka et al., 2013), to partially speciﬁed (Zhu & Mumford, 2006; Roux et al., 2011; Heess et al., 2011; Eslami & Williams, 2014; Tang et al., 2013; 2014), to largely unspeciﬁed (Hinton, 2002; Salakhutdinov & Hin-ton, 2009; Eslami et al., 2012).”
Transitive Invariance for Self-Supervised Visual Representation Learning (2017, methodology): “Earlier methods for generative models include Anto-Encoders [38, 51, 31, 29] and Restricted Boltzmann Machines (RBMs) [23, 4, 49, 12].”
Iterative Instance Segmentation (2015, background): “A related line of work explores learning a generative model of masks [9] using a deep Boltzmann machine [28].”
Temporal Generative Adversarial Nets with Singular Value Clipping (2016, methodology): “A common approach for generating images is the use of undirected graphical models such as Boltzmann machines [31, 18, 4].”
Deep Value Networks Learn to Evaluate and Iteratively Refine Structured Outputs (2017, methodology): “Some methods incorporate higher order potentials in CRFs (Kohli et al., 2009) or model global shape priors with Restricted Boltzmann Machines (Li et al., 2013; Kae et al., 2013; Yang et al., 2014; Eslami et al., 2014).”
Deep learning for visual understanding: A review (2016, background): “The Shape Boltzmann Machine was proposed by Eslami et al. [168] to handle the task of modeling binary shape images, which learns high quality probability distributions over object shapes, for both realism of samples from the distribution and generalization to new examples of the same shape class.”
Modeling object appearance using Context-Conditioned Component Analysis (2015, methodology): “Existing methods for statistical shape modeling, for example the recent ShapeBM model [10], produce generative Figure 1.”
End-to-End Instance Segmentation with Recurrent Attention (2016, background): “[11] proposes a restricted Boltzmann machine to capture high-order pixel interactions for shape modelling.”
Exploring Compositional High Order Pattern Potentials for Structured Output Learning (2013, methodology): “For example, they have been used to model object shape [2], images under occlusion [18], and noisy images [30].”
Deep learning for visual understanding (2017, unknown): “[100] to handle the task of modeling binary shape images, which learns high quality probability distributions over object shapes, for both realism of samples from the distribution and generalization to new examples of the same shape class.”
Learning High-level Prior with Convolutional Neural Networks for Semantic Segmentation (2015, methodology): “The work such as Eslami et al. [3], Yang et al. [36], Kae et al. [14] and Li et al. [20] utilize unsupervised models such as Boltzmann Machine [9], Deep Belief Network [8] and Deep Boltzmann Machine [29] to introduce global constraint for segmentation.”
Hierarchical Convolutional Deep Learning in Computer Vision (2013, methodology): “[32], as applied to images, and the Deep Energy Models of Ngiam et al.”
The informed sampler: A discriminative approach to Bayesian inference in generative computer vision models (2014, methodology): “[6], or it modelled only a limited aspect such as object shape masks [16], or, in the worst case, the generative model was merely used to generate training data for a discriminative model [39].”
3 D ShapeNets : A Deep Representation for Volumetric Shape Modeling (2015, background): “The great generative power of deep learning models has allowed researchers to build deep generative models for 2D shapes: most notably the DBN [14] to generate handwritten digits and ShapeBM [10] to generate horses, etc.”
Augmenting CRFs with Boltzmann Machine Shape Priors for Image Labeling (2013, methodology): “[5] introduced the Shape Boltzmann machine (ShapeBM), a two-layer DBM with local connectivity in the first layer for local consistency and generalization (by weight sharing), and full connectivity in the second layer for modeling global shapes, as a strong model for object shapes.”
A generative restricted Boltzmann machine based method for high-dimensional motion data modeling (2015, methodology): “In literature, RBMs have been successfully applied to separately capture the spatial [9] or temporal [10] patterns in different types of data.”