Associative hierarchical CRFs for object class image segmentation

By Lubor Ladicky, Chris Russell, Pushmeet Kohli, Philip H. S. Torr. 2009

704 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

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs (2016, background): “Even thoughworks of increased sophistication have been proposed to model the hierarchical dependency [26], [27], [28] and/or high-order dependencies”
Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs (2014, background): “Even though works of increased sophistication have been proposed to model the hierarchical dependency (He et al., 2004; Ladicky et al., 2009; Lempitsky et al., 2011) and/or high-order dependencies of segments (Delong et al.”
Multi-Scale Context Aggregation by Dilated Convolutions (2015, background): “Our work shows that the dilated convolution operator is particularly suited to dense prediction due to its ability to expand the receptive ﬁeld without losing resolution or coverage.”
Conditional Random Fields as Recurrent Neural Networks (2015, background): “Lack of such smoothness constraints can result in poor object delineation and small spurious regions in the segmentation output [31, 30, 14, 21].”
Encoder-Decoder with Atrous Separable Convolution for Semantic Image Segmentation (2018, methodology): “There are several model variants proposed to exploit the contextual information for segmentation [28, 65, 36, 39, 22, 79, 51, 14], including those that employ multi-scale inputs (i.”
Rethinking Atrous Convolution for Semantic Image Segmentation (2017, background): “It has been shown that global features or contextual interactions [28, 65, 36, 40, 22, 77] are beneficial in correctly classifying pixels for semantic segmentation.”
Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials (2011, methodology): “[13] by adding color, histogram of oriented gradients (HOG), and pixel location features.”
Feedforward semantic segmentation with zoom-out features (2014, neutral): “Finally, despite the surprising success of the zoom-out architecture described here, we by no means intend to entirely dismiss CRFs, or more generally inference in structured models.”
Auto-DeepLab: Hierarchical Neural Architecture Search for Semantic Image Segmentation (2019, background): “It has been known that context information is crucial for pixel labeling tasks [26, 69, 37, 39, 16, 54, 14, 10].”
Searching for Efficient Multi-Scale Architectures for Dense Image Prediction (2018, background): “A critical element of building such systems is supplying global features and context information to perform pixel-level classiﬁcation [35, 78, 41, 45, 31, 92, 60, 19, 63].”
Higher Order Conditional Random Fields in Deep Neural Networks (2015, methodology): “Conditional Random Fields (CRFs) have been the most common framework, and various rich and expressive models [27,29,48], based on higher order clique potentials, have been developed to improve segmentation performance.”
DeeperLab: Single-Shot Image Parser (2019, background): “For example, it has been known that contextual information is essential for pixel labeling [23, 24, 25, 26, 27, 28, 29].”
Discrete optimization of ray potentials for semantic 3D reconstruction (2015, methodology): “Implementation Details As an input our method uses semantic likelihoods, predicted by a pixel-wise context-based classifier from [15], and depth likelihoods obtained by a plane sweep stereo matching algorithm using zero-mean normalized crosscorrelation.”
Nonparametric semantic segmentation for 3D street scenes (2013, background): “Considerable effort has been focussed towards geometrically modelling the environment with a map consisting of sparse or dense point cloud for accurate localisation [1]–[4].”
ImageNet Auto-Annotation with Segmentation Propagation (2014, methodology): “A simple and widely used technique is to group pixels into superpixels [67,39], and assume that all the pixels inside a superpixel share the same label.”
Diverse M-Best Solutions in Markov Random Fields (2012, methodology): “In a sequence of papers [15, 16], the authors developed a MAP inference algorithm for this model, and we are directly able to utilize this algorithm to compute DivMBest solutions, by rerunning it with modified node energies.”
Semantic Segmentation based on Stacked Discriminative Autoencoders and Context-Constrained Weakly Supervised Learning (2015, methodology): “Table 1 reports the average accuracy over all 21 object classes in comparison with state-of-the-art semantic segmentation methods, including both fully supervised (FS) [10, 11] and weakly supervised (WS) [4, 5].”
Joint Semantic Segmentation and 3D Reconstruction from Monocular Video (2014, methodology): “[20] consisting of color, histogram of oriented gradients (HOG), pixel location features and several filter banks.”
Data-driven fluid simulations using regression forests (2015, methodology): “Various forms of integral features have been previously used for example in the face detector of [Viola and Jones 2004] to calculate Haar wavelets, or in 2D semantic pixelwise classifiers [Shotton et al. 2006; Ladicky et al. 2009] for the fast bag-of-words representation calculation.”