Matches in SemOpenAlex for { <https://semopenalex.org/work/W2982208971> ?p ?o ?g. }
- W2982208971 abstract "Semantic segmentation is a fundamental problem in computer vision. It is considered as a pixel-wise classification problem in practice, and most segmentation models use a pixel-wise loss as their optimization riterion. However, the pixel-wise loss ignores the dependencies between pixels in an image. Several ways to exploit the relationship between pixels have been investigated, eg, conditional random fields (CRF) and pixel affinity based methods. Nevertheless, these methods usually require additional model branches, large extra memories, or more inference time. In this paper, we develop a region mutual information (RMI) loss to model the dependencies among pixels more simply and efficiently. In contrast to the pixel-wise loss which treats the pixels as independent samples, RMI uses one pixel and its neighbour pixels to represent this pixel. Then for each pixel in an image, we get a multi-dimensional point that encodes the relationship between pixels, and the image is cast into a multi-dimensional distribution of these high-dimensional points. The prediction and ground truth thus can achieve high order consistency through maximizing the mutual information (MI) between their multi-dimensional distributions. Moreover, as the actual value of the MI is hard to calculate, we derive a lower bound of the MI and maximize the lower bound to maximize the real value of the MI. RMI only requires a few extra computational resources in the training stage, and there is no overhead during testing. Experimental results demonstrate that RMI can achieve substantial and consistent improvements in performance on PASCAL VOC 2012 and CamVid datasets. The code is available at this https URL." @default.
- W2982208971 created "2019-11-01" @default.
- W2982208971 creator A5003629926 @default.
- W2982208971 creator A5023706427 @default.
- W2982208971 creator A5037942269 @default.
- W2982208971 creator A5046392388 @default.
- W2982208971 date "2019-10-26" @default.
- W2982208971 modified "2023-09-23" @default.
- W2982208971 title "Region Mutual Information Loss for Semantic Segmentation" @default.
- W2982208971 cites W1874027545 @default.
- W2982208971 cites W1903029394 @default.
- W2982208971 cites W2031173376 @default.
- W2982208971 cites W2031489346 @default.
- W2982208971 cites W2034432063 @default.
- W2982208971 cites W2099111195 @default.
- W2982208971 cites W2117539524 @default.
- W2982208971 cites W2120552749 @default.
- W2982208971 cites W2123649031 @default.
- W2982208971 cites W2124592697 @default.
- W2982208971 cites W2133665775 @default.
- W2982208971 cites W2144794286 @default.
- W2982208971 cites W2156875677 @default.
- W2982208971 cites W2160952634 @default.
- W2982208971 cites W2161236525 @default.
- W2982208971 cites W2167843024 @default.
- W2982208971 cites W2171943915 @default.
- W2982208971 cites W2194775991 @default.
- W2982208971 cites W2296319761 @default.
- W2982208971 cites W2412782625 @default.
- W2982208971 cites W2468317448 @default.
- W2982208971 cites W2531409750 @default.
- W2982208971 cites W2557283755 @default.
- W2982208971 cites W2560023338 @default.
- W2982208971 cites W2630837129 @default.
- W2982208971 cites W2737312250 @default.
- W2982208971 cites W2745943519 @default.
- W2982208971 cites W2787091153 @default.
- W2982208971 cites W2789983685 @default.
- W2982208971 cites W2888340395 @default.
- W2982208971 cites W2895420332 @default.
- W2982208971 cites W2899771611 @default.
- W2982208971 cites W2902332991 @default.
- W2982208971 cites W2953384591 @default.
- W2982208971 cites W2962835968 @default.
- W2982208971 cites W2963630186 @default.
- W2982208971 cites W2963739249 @default.
- W2982208971 cites W2963815618 @default.
- W2982208971 cites W2963881378 @default.
- W2982208971 cites W2964242696 @default.
- W2982208971 hasPublicationYear "2019" @default.
- W2982208971 type Work @default.
- W2982208971 sameAs 2982208971 @default.
- W2982208971 citedByCount "3" @default.
- W2982208971 countsByYear W29822089712021 @default.
- W2982208971 crossrefType "posted-content" @default.
- W2982208971 hasAuthorship W2982208971A5003629926 @default.
- W2982208971 hasAuthorship W2982208971A5023706427 @default.
- W2982208971 hasAuthorship W2982208971A5037942269 @default.
- W2982208971 hasAuthorship W2982208971A5046392388 @default.
- W2982208971 hasConcept C105045736 @default.
- W2982208971 hasConcept C152565575 @default.
- W2982208971 hasConcept C153180895 @default.
- W2982208971 hasConcept C154945302 @default.
- W2982208971 hasConcept C160633673 @default.
- W2982208971 hasConcept C2776214188 @default.
- W2982208971 hasConcept C41008148 @default.
- W2982208971 hasConcept C89600930 @default.
- W2982208971 hasConceptScore W2982208971C105045736 @default.
- W2982208971 hasConceptScore W2982208971C152565575 @default.
- W2982208971 hasConceptScore W2982208971C153180895 @default.
- W2982208971 hasConceptScore W2982208971C154945302 @default.
- W2982208971 hasConceptScore W2982208971C160633673 @default.
- W2982208971 hasConceptScore W2982208971C2776214188 @default.
- W2982208971 hasConceptScore W2982208971C41008148 @default.
- W2982208971 hasConceptScore W2982208971C89600930 @default.
- W2982208971 hasLocation W29822089711 @default.
- W2982208971 hasOpenAccess W2982208971 @default.
- W2982208971 hasPrimaryLocation W29822089711 @default.
- W2982208971 hasRelatedWork W114219606 @default.
- W2982208971 hasRelatedWork W1647093459 @default.
- W2982208971 hasRelatedWork W2037559018 @default.
- W2982208971 hasRelatedWork W2067437669 @default.
- W2982208971 hasRelatedWork W2184920060 @default.
- W2982208971 hasRelatedWork W2373817620 @default.
- W2982208971 hasRelatedWork W2377518366 @default.
- W2982208971 hasRelatedWork W2391214433 @default.
- W2982208971 hasRelatedWork W2409973804 @default.
- W2982208971 hasRelatedWork W2546630027 @default.
- W2982208971 hasRelatedWork W2560023338 @default.
- W2982208971 hasRelatedWork W2785464391 @default.
- W2982208971 hasRelatedWork W2810076684 @default.
- W2982208971 hasRelatedWork W3016080479 @default.
- W2982208971 hasRelatedWork W3021698969 @default.
- W2982208971 hasRelatedWork W3132533857 @default.
- W2982208971 hasRelatedWork W3153450104 @default.
- W2982208971 hasRelatedWork W83726181 @default.
- W2982208971 hasRelatedWork W2091458977 @default.
- W2982208971 hasRelatedWork W3177779515 @default.
- W2982208971 isParatext "false" @default.
- W2982208971 isRetracted "false" @default.