Fast Approximate Energy Minimization with Label Costs

By Andrew Delong, Anton Osokin, Hossam N. Isack, Yuri Boykov. 2010

603 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

Semantic Image Segmentation with Deep Convolutional Nets and Fully Connected CRFs (2014, methodology): “…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., 2012; Gonfaus et al., 2010; Kohli et al., 2009; Chen et al., 2013; Wang et al., 2015), we use the fully connected pairwise CRF…”
Weakly-and Semi-Supervised Learning of a Deep Convolutional Network for Semantic Image Segmentation (2015, methodology): “We have experimented with two variants, EM-Fixed and EM-Adapt .”
Object detection by labeling superpixels (2015, methodology): “When the smooth cost term is a metric, the energy function can be solved by the extended α-expansion algorithm with well characterized optimal bounds as proved in [9].”
Harmony Potentials (2011, methodology): “In (Winn and Jojic, 2005), an unsupervised procedure is able to segment an object class using a learned class mask and a deformation ﬁeld.”
Tile-Level Annotation of Satellite Images Using Multi-Level Max-Margin Discriminative Random Field (2013, methodology): “Here we utilize CRF [13,14] to optimize the soft label field, which can directly model the posterior probability of classes.”
Energy-Based Geometric Multi-model Fitting (2012, methodology): “As long as the number of labels is finite (≤10000 or so), such approach can be handled by graphbased optimization methods, e.g. α-expansion (Boykov et al. 2001; Delong et al. 2011).”
On Detection of Multiple Object Instances Using Hough Transforms (2010, background): “image elements (unlike [12], [14], [15]) allows us to retain much of the speed of the traditional Hough transform,”
Efficient Image and Video Co-localization with Frank-Wolfe Algorithm (2014, background): “Similar discrete optimization approaches have been shown to work well in various computer vision applications [6, 12, 13].”
Minimizing Energies with Hierarchical Costs (2012, methodology): “With respect to our energy itself, the most relevant work is the “label costs” of Delong et al. (2012).”
Weakly- and Semi-Supervised Learning of a DCNN for Semantic Image Segmentation (2015, methodology): “image area to be assigned to class l, if z l = 1, and enforce that no pixel is assigned to class l, if z l = 0. We set the parameters ˆ l = ˆ fg, if l>0 and 0 = ˆ bg. Similar constraints appear in [10,20]. In practice, we employ a variant of Algorithm1. We adaptively set the image- and class-depe”
Secrets of GrabCut and Kernel K-Means (2015, background): “If both means and covariances for Gaussians are treated as variables, then (1) corresponds to the standard elliptic K-means energy [31, 29, 10].”
Graph Cut Based Inference with Co-occurrence Statistics (2010, background): “Two efficient schemes [7,12] have been proposed for the minimisation of the number of classes or objects present in a scene.”
UA-DETRAC: A new benchmark and protocol for multi-object detection and tracking (2015, methodology): “[4] extended this approach and proposed to formulate multi-object tracking as a discrete-continuous optimization problem that integrated data association and trajectory estimation into a consistent energy, which was similarly solved by the approach in [16].”
Learning a Dictionary of Shape Epitomes with Applications to Image Labeling (2013, methodology): “This node favors sparsity among global nodes (similar to [4]) by introducing a set of edges {Ega} from {Vg} to Va.”