Exploring context with deep structured models for semantic segmentation

dc.contributor.authorLin, G.
dc.contributor.authorShen, C.
dc.contributor.authorHengel, A.
dc.contributor.authorReid, I.
dc.date.issued2016
dc.descriptionDate of publication 25 May 2017; date of current version 14 May 2018.
dc.description.abstractWe propose an approach for exploiting contextual information in semantic image segmentation, and particularly investigate the use of patch-patch context and patch-background context in deep CNNs. We formulate deep structured models by combining CNNs and Conditional Random Fields (CRFs) for learning the patch-patch context between image regions. Specifically, we formulate CNN-based pairwise potential functions to capture semantic correlations between neighboring patches. Efficient piecewise training of the proposed deep structured model is then applied in order to avoid repeated expensive CRF inference during the course of back propagation. For capturing the patch-background context, we show that a network design with traditional multi-scale image inputs and sliding pyramid pooling is very effective for improving performance.We perform comprehensive evaluation of the proposed method.We achieve new state-of-the-art performance on a number of challenging semantic segmentation datasets.
dc.description.statementofresponsibilityGuosheng Lin, Chunhua Shen, Anton van den Hengel, and Ian Reid
dc.identifier.citationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2016; 40(6):1352-1366
dc.identifier.doi10.1109/TPAMI.2017.2708714
dc.identifier.issn0162-8828
dc.identifier.issn2160-9292
dc.identifier.orcidHengel, A. [0000-0003-3027-8364]
dc.identifier.urihttp://hdl.handle.net/2440/111936
dc.language.isoen
dc.publisherIEEE
dc.relation.granthttp://purl.org/au-research/grants/arc/FT120100969
dc.relation.granthttp://purl.org/au-research/grants/arc/FL130100102
dc.rights© 2017 IEEE
dc.source.urihttps://doi.org/10.1109/tpami.2017.2708714
dc.subjectSemantic segmentation; convolutional neural networks; conditional random fields; contextual models
dc.titleExploring context with deep structured models for semantic segmentation
dc.typeJournal article
pubs.publication-statusPublished

Files