Learning from Weak and Noisy Labels for Semantic Segmentation

Handle URI:
http://hdl.handle.net/10754/608585
Title:
Learning from Weak and Noisy Labels for Semantic Segmentation
Authors:
Lu, Zhiwu; Fu, Zhenyong; Xiang, Tao; Han, Peng; Wang, Liwei; Gao, Xin ( 0000-0002-7108-3574 )
Abstract:
A weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these ‘free’ tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L1-optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L1-optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.
KAUST Department:
Computational Bioscience Research Center (CBRC); Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division
Citation:
Learning from Weak and Noisy Labels for Semantic Segmentation 2016:1 IEEE Transactions on Pattern Analysis and Machine Intelligence
Publisher:
Institute of Electrical and Electronics Engineers (IEEE)
Journal:
IEEE Transactions on Pattern Analysis and Machine Intelligence
Issue Date:
8-Apr-2016
DOI:
10.1109/TPAMI.2016.2552172
Type:
Article
ISSN:
0162-8828; 2160-9292
Sponsors:
This work was partially supported by National Natural Science Foundation of China (61573363 and 61573026), 973 Program of China (2014CB340403 and 2015CB352502), the Fundamental Research Funds for the Central Universities and the Research Funds of Renmin University of China (15XNLQ01), IBM Global SUR Award Program, European Research Council FP7 Project SUNNY (313243), and the funding from KAUST.
Additional Links:
http://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7450177
Appears in Collections:
Articles; Computational Bioscience Research Center (CBRC); Computer, Electrical and Mathematical Sciences and Engineering (CEMSE) Division

Full metadata record

DC FieldValue Language
dc.contributor.authorLu, Zhiwuen
dc.contributor.authorFu, Zhenyongen
dc.contributor.authorXiang, Taoen
dc.contributor.authorHan, Pengen
dc.contributor.authorWang, Liweien
dc.contributor.authorGao, Xinen
dc.date.accessioned2016-05-08T14:08:45Zen
dc.date.available2016-05-08T14:08:45Zen
dc.date.issued2016-04-08en
dc.identifier.citationLearning from Weak and Noisy Labels for Semantic Segmentation 2016:1 IEEE Transactions on Pattern Analysis and Machine Intelligenceen
dc.identifier.issn0162-8828en
dc.identifier.issn2160-9292en
dc.identifier.doi10.1109/TPAMI.2016.2552172en
dc.identifier.urihttp://hdl.handle.net/10754/608585en
dc.description.abstractA weakly supervised semantic segmentation (WSSS) method aims to learn a segmentation model from weak (image-level) as opposed to strong (pixel-level) labels. By avoiding the tedious pixel-level annotation process, it can exploit the unlimited supply of user-tagged images from media-sharing sites such as Flickr for large scale applications. However, these ‘free’ tags/labels are often noisy and few existing works address the problem of learning with both weak and noisy labels. In this work, we cast the WSSS problem into a label noise reduction problem. Specifically, after segmenting each image into a set of superpixels, the weak and potentially noisy image-level labels are propagated to the superpixel level resulting in highly noisy labels; the key to semantic segmentation is thus to identify and correct the superpixel noisy labels. To this end, a novel L1-optimisation based sparse learning model is formulated to directly and explicitly detect noisy labels. To solve the L1-optimisation problem, we further develop an efficient learning algorithm by introducing an intermediate labelling variable. Extensive experiments on three benchmark datasets show that our method yields state-of-the-art results given noise-free labels, whilst significantly outperforming the existing methods when the weak labels are also noisy.en
dc.description.sponsorshipThis work was partially supported by National Natural Science Foundation of China (61573363 and 61573026), 973 Program of China (2014CB340403 and 2015CB352502), the Fundamental Research Funds for the Central Universities and the Research Funds of Renmin University of China (15XNLQ01), IBM Global SUR Award Program, European Research Council FP7 Project SUNNY (313243), and the funding from KAUST.en
dc.language.isoenen
dc.publisherInstitute of Electrical and Electronics Engineers (IEEE)en
dc.relation.urlhttp://ieeexplore.ieee.org/lpdocs/epic03/wrapper.htm?arnumber=7450177en
dc.rights(c) 2016 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other users, including reprinting/ republishing this material for advertising or promotional purposes, creating new collective works for resale or redistribution to servers or lists, or reuse of any copyrighted components of this work in other works.en
dc.subjectSemantic segmentationen
dc.subjectlabel noise reductionen
dc.subjectsparse learningen
dc.subjectweakly supervised learningen
dc.titleLearning from Weak and Noisy Labels for Semantic Segmentationen
dc.typeArticleen
dc.contributor.departmentComputational Bioscience Research Center (CBRC)en
dc.contributor.departmentComputer, Electrical and Mathematical Sciences and Engineering (CEMSE) Divisionen
dc.identifier.journalIEEE Transactions on Pattern Analysis and Machine Intelligenceen
dc.eprint.versionPost-printen
dc.contributor.institutionBeijing Key Laboratory of Big Data Man- agement and Analysis Methods, School of Information, Renmin University of China, Beijing 100872, Chinaen
dc.contributor.institutionSchool of Electronic Engineering and Computer Science, Queen Mary University of London, Mile End Road, London E1 4NS, United Kingdomen
dc.contributor.institutionSchool of Electronics Engineering and Computer Science, Peking University, Beijing 100871, Chinaen
dc.contributor.affiliationKing Abdullah University of Science and Technology (KAUST)en
kaust.authorGao, Xinen
All Items in KAUST are protected by copyright, with all rights reserved, unless otherwise indicated.