A Fast, Massively Parallel Solver for Large, Irregular Pairwise Markov Random Fields

dc.contributor.authorThuerck, Danielen_US
dc.contributor.authorWaechter, Michaelen_US
dc.contributor.authorWidmer, Svenen_US
dc.contributor.authorBuelow, Max vonen_US
dc.contributor.authorSeemann, Patricken_US
dc.contributor.authorPfetsch, Marc E.en_US
dc.contributor.authorGoesele, Michaelen_US
dc.contributor.editorUlf Assarsson and Warren Hunten_US
dc.date.accessioned2016-06-17T14:08:32Z
dc.date.available2016-06-17T14:08:32Z
dc.date.issued2016en_US
dc.description.abstractGiven the increasing availability of high-resolution input data, today's computer vision problems tend to grow beyond what has been considered tractable in the past. This is especially true for Markov Random Fields (MRFs), which have expanded beyond millions of variables with thousands of labels. Such MRFs pose new challenges for inference, requiring massively parallel solvers that can cope with large-scale problems and support general, irregular input graphs. We propose a block coordinate descent based solver for large MRFs designed to exploit many-core hardware such as recent GPUs. We identify tree-shaped subgraphs as a block coordinate scheme for irregular topologies and optimize them efficiently using dynamic programming. The resulting solver supports arbitrary MRF topologies efficiently and can handle arbitrary, dense or sparse label sets as well as label cost functions. Together with two additional heuristics for further acceleration, our solver performs favorably even compared to modern specialized solvers in terms of speed and solution quality, especially when solving very large MRFs.en_US
dc.description.sectionheadersVR and GPU Computeen_US
dc.description.seriesinformationEurographics/ ACM SIGGRAPH Symposium on High Performance Graphicsen_US
dc.identifier.doi10.2312/hpg.20161203en_US
dc.identifier.isbn978-3-03868-008-6en_US
dc.identifier.issn2079-8679en_US
dc.identifier.pages173-183en_US
dc.identifier.urihttps://doi.org/10.2312/hpg.20161203en_US
dc.publisherThe Eurographics Associationen_US
dc.subjectI.4.m [Image Processing and Computer Vision]en_US
dc.subjectMiscellaneousen_US
dc.subjectProbabilistic Modelsen_US
dc.titleA Fast, Massively Parallel Solver for Large, Irregular Pairwise Markov Random Fieldsen_US
Files
Original bundle
Now showing 1 - 2 of 2
Loading...
Thumbnail Image
Name:
173-183.pdf
Size:
3.11 MB
Format:
Adobe Portable Document Format
Loading...
Thumbnail Image
Name:
mapmap_supplemental_material.pdf
Size:
3.82 MB
Format:
Adobe Portable Document Format