Logo des Repositoriums
 

Active-learning-driven deep interactive segmentation for cost-effective labeling of crop-weed image data

dc.contributor.authorSikouonmeu, Freddy
dc.contributor.authorAtzmueller, Martin
dc.contributor.editorHoffmann, Christa
dc.contributor.editorStein, Anthony
dc.contributor.editorRuckelshausen, Arno
dc.contributor.editorMüller, Henning
dc.contributor.editorSteckel, Thilo
dc.contributor.editorFloto, Helga
dc.date.accessioned2023-02-21T15:14:20Z
dc.date.available2023-02-21T15:14:20Z
dc.date.issued2023
dc.description.abstractActive learning has shown its reliability in (semi-)supervised machine learning tasks to reduce the labeling cost for large datasets in various areas. However, in the agricultural field, despite past attempts to reduce the labeling cost and the burden on the labeler in acquiring image labels, the load during the acquisition of pixel-level labels for semantic image segmentation tasks remains high. Typically, the respective pixel-level masks are acquired manually by drawing polygons over irregular and complex-shaped object boundaries. In contrast, this paper proposes a method leveraging the power of a click-based deep interactive segmentation model (DISEG) in an active learning approach to harvest high-quality image segmentation labels at a low cost for training a real-time task model by only clicking on the objects’ fore- and background surfaces. Our first experimental results indicate that with an average of 3 clicks per image object and using only 3% of the unlabeled dataset, we can acquire pixel-level labels with good quality at low cost.en
dc.identifier.isbn978-3-88579-724-1
dc.identifier.pissn1617-5468
dc.identifier.urihttps://dl.gi.de/handle/20.500.12116/40301
dc.language.isoen
dc.publisherGesellschaft für Informatik e.V.
dc.relation.ispartof43. GIL-Jahrestagung, Resiliente Agri-Food-Systeme
dc.relation.ispartofseriesLecture Notes in Informatics (LNI) - Proceedings, Volume P-330
dc.subjectdeep learning
dc.subjectactive learning
dc.subjectdeep interactive segmentation
dc.subjectreal-time semantic segmentation
dc.subjectcrop-weed detection
dc.subjectdata annotation
dc.titleActive-learning-driven deep interactive segmentation for cost-effective labeling of crop-weed image dataen
dc.typeText/Conference Paper
gi.citation.endPage512
gi.citation.publisherPlaceBonn
gi.citation.startPage507
gi.conference.date13.-14. Februar 2023
gi.conference.locationOsnabrück

Dateien

Originalbündel
1 - 1 von 1
Lade...
Vorschaubild
Name:
GIL_2023_Sikouonmeu_507-512.pdf
Größe:
415.97 KB
Format:
Adobe Portable Document Format