Scene Segmentation with CRFs Learned from Partially Labeled Images

By Jakob Verbeek, Bill Triggs. 2007

150 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

Diffusion-Convolutional Neural Networks (2015, methodology): “observed CRF is usually computed using a contrast-of-partition-functions approach that requires running loopy belief propagation twice; once on the entire graph and once with the observed labels ﬁxed [14]. This algorithm, and thus each step in the numerical optimization, has exponential time complex”
Efficient Inference in Fully Connected CRFs with Gaussian Edge Potentials (2011, background): “Basic CRF models are composed of unary potentials on individual pixels or image patches and pairwise potentials on neighboring pixels or patches [19, 23, 7, 5].”
Harmony potentials for joint classification and segmentation (2010, methodology): “Verbeek and Triggs [22] proposed to use global scale information to improve estimation at local scales.”
Seed, Expand and Constrain: Three Principles for Weakly-Supervised Image Segmentation (2016, methodology): “were developed that only require partial segmentations [36,11] or object bounding boxes [19,51,8] as training data.”
Context based object categorization: A critical survey (2010, unknown): “Region level interactions have been extensively investigated in the area of context-based object categorization tasks [6,7,15,16,19,20,27,31,37] since regions follow plausible geometrical configurations.”
Object categorization using co-occurrence, location and appearance (2008, background): “Along these lines, other approaches have also considered global image features as a source of context; either by using the correlation of low level features across images that contain the object or across the category [8, 14, 23, 27, 29].”
Learning Graphical Model Parameters with Approximate Marginal Inference (2013, methodology): “However, using the same approximation for both Að Þ and Að ; zÞ appears to work well in practice [26].”
Recurrent Convolutional Neural Networks for Scene Labeling (2014, methodology): “These models are then trained to maximize the likelihood of correct classification given the features [1, 2, 3, 4, 5, 6, 7].”