Learning to Combine Bottom-Up and Top-Down Segmentation

By Anat Levin, Yair Weiss. International Journal of Computer Vision, 2006

320 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Genetics, Radiation

View Paper: https://www.semanticscholar.org/paper/26c8c478a63b880cc52634b72e4b24e885269511

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 Genetics

In Radiation

Citations in the corpus

Detecting Object Boundaries Using Low-, Mid-, and High-level Information (2007, methodology): “Combining both top-down and bottomup learning in a loop is emphasized in [20].”
What Happened to My Dog in That Network: Unraveling Top-down Generators in Convolutional Neural Networks (2015, background): “The chain of research stretches back even to the early days of computer vision research, but more recent works demonstrating the performance benefits of top-down information in tasks such as object perception include (Borenstein & Ullman, 2008; Tu et al., 2005; Levin & Weiss, 2009).”
Scene Text Recognition using Higher Order Language Priors (2009, methodology): “Other methods address the challenging task of combined object recognition and pixel-wise segmentation [5, 13, 17, 21].”
Crisp Boundary Detection Using Pointwise Mutual Information (2014, methodology): “Object boundary cues have been used to aid in segmentation [1,2,3], object detection and recognition [4,5], and recovery of intrinsic scene properties such as shape, reflectance, and illumination [6].”
Learning Graphical Model Parameters with Approximate Marginal Inference (2013, methodology): “This “surrogate likelihood” [17] approximation appears to be the most widely used loss in imaging problems, with marginals approximated by either MNF [18], [19], TRW [20], or LBP [21], [22], [23], [24], [25].”
The truth about cats and dogs (2011, background): “However, often the goal of these methods has been segmentation of the entire image, rather than object category detection, whilst others [2, 3, 20, 21, 27, 28, 37] have generally targeted typical views of vehicles and animals (e.”
Harmony Potentials (2011, methodology): “It is usually used by top-down object segmentation approaches, which use the mid-level context scale to disambiguate local predictions and, in contrast to bottom-up approaches, they use a priori knowledge about the whole object such as its structure (Levin and Weiss, 2009).”
Figure-ground segmentation by transferring window masks (2012, methodology): “In this work instead we estimate the appearance models automatically, based on the segmentation transfer idea.”
Delving deeper into the whorl of flower segmentation (2010, result): “Similar ideas are present in several other recent class based segmentation methods [3–6].”
Object Recognition by Sequential Figure-Ground Ranking (2011, background): “Interesting solutions have been proposed recently, although generalization to a large class of shapes remains non-trivial (Kumar et al. 2005; Levin and Weiss 2009).”
Layered object detection for multi-class segmentation (2010, unknown): “Early approaches bias a segmentation engine using the output of object models [27, 17], while others iterate between bottom-up and top-down cues [14, 15, 23].”
A Primal-Dual Message-Passing Algorithm for Approximated Large Scale Structured Prediction (2010, background): “[2] for structured SVMs and [20, 12] for CRFs.”
CRF Learning with CNN Features for Image Segmentation (2015, methodology): “Method Sa So Levin & Weiss [27] 95.”
CPMC: Automatic Object Segmentation Using Constrained Parametric Min-Cuts (2012, unknown): “Expressive part-based appearance models have also been developed [49], [50], [51], [52].”