Discriminative random fields: a discriminative framework for contextual interaction in classification

By Sanjiv Kumar, Martial Hebert. Proceedings Ninth IEEE International Conference on Computer Vision, 2003

533 citations recorded by the source corpus.

In this work we present discriminative random fields (DRFs), a discriminative framework for the classification of image regions by incorporating neighborhood interactions in the labels as well as the observed data. The discriminative random fields offer several advantages over the conventional Markov random field (MRF) framework. First, the DRFs allow to relax the strong assumption of conditional independence of the observed data generally used in the MRF framework for tractability. This assumption is too restrictive for a large number of applications in vision. Second, the DRFs derive their classification power by exploiting the probabilistic discriminative models instead of the generative models used in the MRF framework. Finally, all the parameters in the DRF model are estimated simultaneously from the training data unlike the MRF framework where likelihood parameters are usually learned separately from the field parameters. We illustrate the advantages of the DRFs over the MRF framework in an application of man-made structure detection in natural images taken from the Corel database.

Subfields: Computer Vision and Pattern Recognition, Media Technology

View Paper: https://www.semanticscholar.org/paper/e54f01884e1fba4a0bbd2f0989ad21a16ebb13e3

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 Media Technology

Citations in the corpus

Conditional random fields for activity recognition (2007, background): “[2, 11, 14], while conditional random fields, a more recent model, are just now expanding outside of their original natural language processing context into a range of areas such as image processing [4], gesture recognition [13], motion tracking [10], and activity recognition [6].”
TextonBoost for Image Understanding: Multi-Class Object Recognition and Segmentation by Jointly Modeling Texture, Layout, and Context (2007, methodology): “Conversely, other authors have considered only the segmentation task, for example (Kumar and Hebert 2003; Borenstein et al. 2004).”
Location-Based Activity Recognition using Relational Markov Networks (2005, methodology): “Then we show how to use them for modeling activities.”
Learning 3D mesh segmentation and labeling (2010, methodology): “Next, we add a pairwise term to refine these boundaries.”
Discriminative Random Fields (2006, methodology): “These models have been used with some success in various segmentation and labeling problems (Bouman and Shapiro, 1994; Cheng and Bouman, 2001; Feng et al., 2002; Wilson and Li, 2003; Kumar and Hebert, 2003c).”
Multi-View Stereo: A Tutorial (2015, background): “The texture component of the feature vector consists of the statistics of the edge orientation histogram, in particular, entropy, maximum value, and number of modes [125].”
Scale-invariant contour completion using conditional random fields (2005, methodology): “Conditional random fields have been proposed as a method for image segmentation by[12, 23, 10], however, the focus is still on pixel-level labeling which is tied to the resolution of the pixel grid.”
Detecting Object Boundaries Using Low-, Mid-, and High-level Information (2007, methodology): “Conditional Markov random fields models [10, 16] have been used to enforce local consistency between neighboring structures.”
Efficient Multilevel Brain Tumor Segmentation With Integrated Bayesian Model Classification (2008, unknown): “[14] that uses the context-sensitive discriminative random fields model [19], [20].”
Objects in Context (2007, methodology): “CRFs have been widely used in object detection, labeling, and classification [9, 10, 14, 22].”
Auto-Context and Its Application to High-Level Vision Tasks and 3D Brain Image Segmentation (2010, methodology): “In vision, models like Markov Random Fields (MRFs) [13] and Conditional Random Fields (CRFs) [23], [21] have been used to capture the context information.”
Putting Objects in Perspective (2006, background): “lationships between objects and other objects [15, 18], regions [10, 16, 28] or scene categories [18, 24], all within the 2D image.”
RGB-(D) scene labeling: Features and algorithms (2012, background): “He et al. proposed multi-scale CRFs for learning label patterns [10].”
Protein Fold Recognition Using Segmentation Conditional Random Fields (SCRFs) (2006, unknown): “The conditional graphical models prove to be one of the most effective tools for this kind of problem (Kumar and Hebert, 2003; Pinto et al. , 2003).”
Maximin affinity learning of image segmentation (2009, background): “In the cases where probabilistic random field models have been used for image parsing and segmentation, the models have either been simplistic for tractability reasons [12] or have been trained piecemeal.”