CopyRight 2012-2014 DS文库版权所有
U-Net
(0 次评价)1737 人阅读0 次下载

 U-Net:ConvolutionalNetworksforBiomedicalImageSegmentation OlafRonneberger,PhilippFischer,andThomasBrox ComputerScienceDepartmentandBIOSSCentreforBiologicalSignallingStudies,UniversityofFreiburg,Germany ronneber@informatik.uni-freiburg.de ,WWWhomepage: http://lmb.informatik.uni-freiburg.de/ Abstract. Thereislargeconsentthatsuccessfultrainingofdeepnet-worksrequiresmanythousandannotatedtrainingsamples.Inthispa-per,wepresentanetworkandtrainingstrategythatreliesonthestronguseofdataaugmentationtousetheavailableannotatedsamplesmoreefficiently.Thearchitectureconsistsofacontractingpathtocapturecontextandasymmetricexpandingpaththatenablespreciselocaliza-tion.Weshowthatsuchanetworkcanbetrainedend-to-endfromveryfewimagesandoutperformsthepriorbestmethod(asliding-windowconvolutionalnetwork)ontheISBIchallengeforsegmentationofneu-ronalstructuresinelectronmicroscopicstacks.Usingthesamenet-worktrainedontransmittedlightmicroscopyimages(phasecontrastandDIC)wewontheISBIcelltrackingchallenge2015inthesecate-goriesbyalargemargin.Moreover,thenetworkisfast.Segmentationofa512x512imagetakeslessthanasecondonarecentGPU.Thefullimplementation(basedonCaffe)andthetrainednetworksareavailableat http://lmb.informatik.uni-freiburg.de/people/ronneber/u-net . 1Introduction Inthelasttwoyears,deepconvolutionalnetworkshaveoutperformedthestateoftheartinmanyvisualrecognitiontasks,e.g.[ 7 , 3 ].Whileconvolutionalnetworkshavealreadyexistedforalongtime[ 8 ],theirsuccesswaslimitedduetothesizeoftheavailabletrainingsetsandthesizeoftheconsiderednetworks.ThebreakthroughbyKrizhevskyetal.[ 7 ]wasduetosupervisedtrainingofalargenetworkwith8layersandmillionsofparametersontheImageNetdatasetwith1milliontrainingimages.Sincethen,evenlargeranddeepernetworkshavebeentrained[ 12 ].Thetypicaluseofconvolutionalnetworksisonclassificationtasks,wheretheoutputtoanimageisasingleclasslabel.However,inmanyvisualtasks,especiallyinbiomedicalimageprocessing,thedesiredoutputshouldincludelocalization,i.e.,aclasslabelissupposedtobeassignedtoeachpixel.More-over,thousandsoftrainingimagesareusuallybeyondreachinbiomedicaltasks.Hence,Ciresanetal.[ 1 ]trainedanetworkinasliding-windowsetuptopredicttheclasslabelofeachpixelbyprovidingalocalregion(patch)aroundthatpixel arXiv:1505.04597v1 [cs.CV] 18 May 2015

打分:

0 星

用户评论:

小飞飞
于 2020-07-15 上传

版权及免责声明|RISC-V单片机中文网 |网站地图

GMT+8, 2022-10-2 14:06 , Processed in 0.763125 second(s), 30 queries .

返回顶部