Learning CRFs Using Graph Cuts

By Martin Szummer, Pushmeet Kohli, Derek Hoiem. European Conference on Computer Vision, 2008

303 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/eb86d3cd5d18477be5164b5c9335bfb79e7124a7

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

Deep Learning and Structured Prediction for the Segmentation of Mass in Mammograms (2015, methodology): “(4), where the introduction of slack variable leads to [9, 10]:”
Learning Efficient Random Maximum A-Posteriori Predictors with Non-Decomposable Loss Functions (2013, methodology): “In our experiments the structured SVM with the hamming loss did not perform well – the loss augmented inference tended to focus on maximum violations instead of good solutions which causes the parameters to change even though the MAP solution has a low loss (a similar phenomenon was observed in [22]”
Pylon Model for Semantic Segmentation (2011, background): “(2) Once again, the constraint (2) is not needed for flat CRFs as their pools do not contain overlapping segments.”
ImageNet Auto-Annotation with Segmentation Propagation (2014, background): “Since ∆ i can be expressed as a unary potential over pixels, this problem can be solved exactly using graph-cut [61].”
Category Independent Object Proposals (2010, methodology): “Learning: To solve the structured learning problem, we use the slack-rescaled method with loss penalty used in [30].”
Contextual classification with functional Max-Margin Markov Networks (2009, unknown): “[24] and Franc and Savchynskyy [5] investigate optimizing the pairwise M3N criterion using cutting-plane techniques which require the use of a quadratic program solver.”
Geodesic star convexity for interactive image segmentation (2010, methodology): “A similar cost function (but without the star constraint), was minimized for the purpose of parameter learning in Szummer et al. [22].”
Top-down visual saliency via joint CRF and dictionary learning (2012, methodology): “CRFs have been demonstrated as a flexible framework of incorporating different kinds of features for visual recognition [25, 24, 5, 7, 1].”
Deep-learning-based object-level contour detection with CCG and CRF optimization (2017, methodology): “Then the SSVM [12] is utilized to obtain the parameter w.”
Learning to localize detected objects (2012, methodology): “Other works exist in structural learning of image segmentation [1], [22], and we follow [22] which learns parameters for different potentials and solves image segmentation using graph cuts.”
Automatic Feature Learning for Robust Shadow Detection (2014, methodology): “16 are learnt using a max-margin criterion, details of which are given in [26].”
SKYLINE2GPS: Localization in urban canyons using omni-skylines (2010, methodology): “Our method is similar to the maximum-margin network learning method using graph cuts [30]; we generate “near-miss” labelings, then estimate a parameter vector that maximizes the margin separating true labellings from the near misses in a convex optimization.”
Weakly Supervised Localization and Learning with Generic Knowledge (2012, unknown): “We learn α according to a max-margin criterion following the constraint-generation approach used to train structured output SVMs [Tsochantaridis et al, 2005], analogously to other work on learning the parameters of a CRF [Deselaers and Ferrari, 2010, Fin-ley and Joachims, 2008, Szummer et al, 2008].”
Semantic Image Segmentation via Deep Parsing Network (2015, background): “Existing works such as [31, 29, 9, 34, 11, 2, 8, 25, 22] can be generally categorized into two groups based on their definitions of the unary and pairwise terms of MRF.”
Hidden Part Models for Human Action Recognition: Probabilistic versus Max Margin (2011, methodology): “One solution is to use a cutting plane algorithm similar to that used in structured SVMs [30] and CRFs [36].”
Segmentation Propagation in ImageNet (2012, unknown): “Our work weaves together and extends several recent developments including Grabcut [1], segmentation transfer [11, 2], efficient binary codes [8], cosegmentation [3, 12] and structured output learning [13, 14] into a fully automatic, computationally efficient and reliable large scale segmentation fr”
Perceptually Inspired Layout-Aware Losses for Image Segmentation (2014, background): “This technique has gained wide-spread acceptance in Computer Vision by producing impressive results for many vision problems including image segmentation [18,21] and object detection [3,6].”