FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W2804077308> ?p ?o ?g. }

• W2804077308 abstract "Dense conditional random fields (CRFs) have become a popular framework for modelling several problems in computer vision such as stereo correspondence and multi-class semantic segmentation. By modelling long-range interactions, dense CRFs provide a labelling that captures finer detail than their sparse counterparts. Currently, the state-of-the-art algorithm performs mean-field inference using a filter-based method but fails to provide a strong theoretical guarantee on the quality of the solution. A question naturally arises as to whether it is possible to obtain a maximum a posteriori (MAP) estimate of a dense CRF using a principled method. Within this paper, we show that this is indeed possible. We will show that, by using a filter-based method, continuous relaxations of the MAP problem can be optimised efficiently using state-of-the-art algorithms. Specifically, we will solve a quadratic programming (QP) relaxation using the Frank-Wolfe algorithm and a linear programming (LP) relaxation by developing a proximal minimisation framework. By exploiting labelling consistency in the higher-order potentials and utilising the filter-based method, we are able to formulate the above algorithms such that each iteration has a complexity linear in the number of classes and random variables. The presented algorithms can be applied to any labelling problem using a dense CRF with sparse higher-order potentials. In this paper, we use semantic segmentation as an example application as it demonstrates the ability of the algorithm to scale to dense CRFs with large dimensions. We perform experiments on the Pascal dataset to indicate that the presented algorithms are able to attain lower energies than the mean-field inference method." @default.
• W2804077308 created "2018-06-01" @default.
• W2804077308 creator A5004223779 @default.
• W2804077308 creator A5013834379 @default.
• W2804077308 creator A5022356850 @default.
• W2804077308 creator A5027893749 @default.
• W2804077308 creator A5039055995 @default.
• W2804077308 creator A5042899882 @default.
• W2804077308 creator A5049300388 @default.
• W2804077308 creator A5078217575 @default.
• W2804077308 date "2018-05-23" @default.
• W2804077308 modified "2023-10-02" @default.
• W2804077308 title "Efficient Relaxations for Dense CRFs with Sparse Higher Order Potentials" @default.
• W2804077308 cites W1511986666 @default.
• W2804077308 cites W1903029394 @default.
• W2804077308 cites W1964669037 @default.
• W2804077308 cites W1964772475 @default.
• W2804077308 cites W2050159143 @default.
• W2804077308 cites W2054279472 @default.
• W2804077308 cites W2061130579 @default.
• W2804077308 cites W2067191022 @default.
• W2804077308 cites W2086225343 @default.
• W2804077308 cites W2102492119 @default.
• W2804077308 cites W2124592697 @default.
• W2804077308 cites W2127847374 @default.
• W2804077308 cites W2128680590 @default.
• W2804077308 cites W2130517245 @default.
• W2804077308 cites W2131241448 @default.
• W2804077308 cites W2136885855 @default.
• W2804077308 cites W2140666719 @default.
• W2804077308 cites W2161236525 @default.
• W2804077308 cites W2252343674 @default.
• W2804077308 cites W2283104426 @default.
• W2804077308 cites W2460087882 @default.
• W2804077308 cites W2517928844 @default.
• W2804077308 cites W2535516436 @default.
• W2804077308 cites W2558109362 @default.
• W2804077308 cites W27675589 @default.
• W2804077308 cites W2776638780 @default.
• W2804077308 cites W2913535645 @default.
• W2804077308 cites W2964177174 @default.
• W2804077308 cites W2964262905 @default.
• W2804077308 cites W2964288706 @default.
• W2804077308 hasPublicationYear "2018" @default.
• W2804077308 type Work @default.
• W2804077308 sameAs 2804077308 @default.
• W2804077308 citedByCount "0" @default.
• W2804077308 crossrefType "proceedings-article" @default.
• W2804077308 hasAuthorship W2804077308A5004223779 @default.
• W2804077308 hasAuthorship W2804077308A5013834379 @default.
• W2804077308 hasAuthorship W2804077308A5022356850 @default.
• W2804077308 hasAuthorship W2804077308A5027893749 @default.
• W2804077308 hasAuthorship W2804077308A5039055995 @default.
• W2804077308 hasAuthorship W2804077308A5042899882 @default.
• W2804077308 hasAuthorship W2804077308A5049300388 @default.
• W2804077308 hasAuthorship W2804077308A5078217575 @default.
• W2804077308 hasConcept C106131492 @default.
• W2804077308 hasConcept C11413529 @default.
• W2804077308 hasConcept C152565575 @default.
• W2804077308 hasConcept C153180895 @default.
• W2804077308 hasConcept C154945302 @default.
• W2804077308 hasConcept C2775953691 @default.
• W2804077308 hasConcept C2776214188 @default.
• W2804077308 hasConcept C31972630 @default.
• W2804077308 hasConcept C41008148 @default.
• W2804077308 hasConceptScore W2804077308C106131492 @default.
• W2804077308 hasConceptScore W2804077308C11413529 @default.
• W2804077308 hasConceptScore W2804077308C152565575 @default.
• W2804077308 hasConceptScore W2804077308C153180895 @default.
• W2804077308 hasConceptScore W2804077308C154945302 @default.
• W2804077308 hasConceptScore W2804077308C2775953691 @default.
• W2804077308 hasConceptScore W2804077308C2776214188 @default.
• W2804077308 hasConceptScore W2804077308C31972630 @default.
• W2804077308 hasConceptScore W2804077308C41008148 @default.
• W2804077308 hasLocation W28040773081 @default.
• W2804077308 hasOpenAccess W2804077308 @default.
• W2804077308 hasPrimaryLocation W28040773081 @default.
• W2804077308 hasRelatedWork W1559345028 @default.
• W2804077308 hasRelatedWork W159509105 @default.
• W2804077308 hasRelatedWork W1602118253 @default.
• W2804077308 hasRelatedWork W1614580364 @default.
• W2804077308 hasRelatedWork W1675973454 @default.
• W2804077308 hasRelatedWork W173075451 @default.
• W2804077308 hasRelatedWork W2065540576 @default.
• W2804077308 hasRelatedWork W2169753597 @default.
• W2804077308 hasRelatedWork W2187697692 @default.
• W2804077308 hasRelatedWork W2435789594 @default.
• W2804077308 hasRelatedWork W2530114498 @default.
• W2804077308 hasRelatedWork W2540616960 @default.
• W2804077308 hasRelatedWork W270386715 @default.
• W2804077308 hasRelatedWork W2951934896 @default.
• W2804077308 hasRelatedWork W2954279345 @default.
• W2804077308 hasRelatedWork W2963062669 @default.
• W2804077308 hasRelatedWork W2999688250 @default.
• W2804077308 hasRelatedWork W3009577965 @default.
• W2804077308 hasRelatedWork W3113297335 @default.
• W2804077308 hasRelatedWork W3209312884 @default.
• W2804077308 isParatext "false" @default.
• W2804077308 isRetracted "false" @default.
• W2804077308 magId "2804077308" @default.

• next