Publikation: Accelerating Active Learning Image Labeling Through Bulk Shift Recommendations
Dateien
Datum
Autor:innen
Herausgeber:innen
ISSN der Zeitschrift
Electronic ISSN
ISBN
Bibliografische Daten
Verlag
Schriftenreihe
Auflagebezeichnung
DOI (zitierfähiger Link)
Internationale Patentnummer
Angaben zur Forschungsförderung
Projekt
Open Access-Veröffentlichung
Core Facility der Universität Konstanz
Titel in einer weiteren Sprache
Publikationstyp
Publikationsstatus
Erschienen in
Zusammenfassung
Nowadays, the inexpensive memory space promotes an accelerating growth of stored image data. To exploit the data using supervised Machine or Deep Learning, it needs to be labeled. Manually labeling the vast amount of data is time-consuming and expensive, especially if human experts with specific domain knowledge are indispensable. Active learning addresses this shortcoming by querying the user the labels of the most informative images first. One way to obtain the ‘informativeness’ is by using uncertainty sampling as a query strategy, where the system queries those images it is most uncertain about how to classify. In this paper, we present a web-based active learning framework that helps to accelerate the labeling process. After manually labeling some images, the user gets recommendations of further candidates that could potentially be labeled equally (bulk image folder shift). We aim to explore the most efficient ‘uncertainty’ measure to improve the quality of the recommendations such that all images are sorted with a minimum number of user interactions (clicks). We conducted experiments using a manually labeled reference dataset to evaluate different combinations of classifiers and uncertainty measures. The results clearly show the effectiveness of an uncertainty sampling with bulk image shift recommendations (our novel method), which can reduce the number of required clicks to only around 20% compared to manual labeling.
Zusammenfassung in einer weiteren Sprache
Fachgebiet (DDC)
Schlagwörter
Konferenz
Rezension
Zitieren
ISO 690
SCHARPF, Philipp, Chi Lap HONG, Oliver DUERR, 2021. Accelerating Active Learning Image Labeling Through Bulk Shift Recommendations. 2021 International Conference on Data Mining Workshops (ICDMW). Virtual Conference, 7. Dez. 2021 - 10. Dez. 2021. In: XUE, Bing, ed., Mykola PECHENIZKIY, ed., Yun SING KOH, ed.. 21th IEEE International Conference on Data Mining workshops : 7-10 December 2021, virtual conference : proceedings. Los Alamitos, et al.: IEEE, 2021, pp. 398-404. ISBN 978-1-66542-427-1. Available under: doi: 10.1109/ICDMW53433.2021.00055BibTex
@inproceedings{Scharpf2021Accel-57373, year={2021}, doi={10.1109/ICDMW53433.2021.00055}, title={Accelerating Active Learning Image Labeling Through Bulk Shift Recommendations}, isbn={978-1-66542-427-1}, publisher={IEEE}, address={Los Alamitos, et al.}, booktitle={21th IEEE International Conference on Data Mining workshops : 7-10 December 2021, virtual conference : proceedings}, pages={398--404}, editor={Xue, Bing and Pechenizkiy, Mykola and Sing Koh, Yun}, author={Scharpf, Philipp and Hong, Chi Lap and Duerr, Oliver} }
RDF
<rdf:RDF xmlns:dcterms="http://purl.org/dc/terms/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#" xmlns:bibo="http://purl.org/ontology/bibo/" xmlns:dspace="http://digital-repositories.org/ontologies/dspace/0.1.0#" xmlns:foaf="http://xmlns.com/foaf/0.1/" xmlns:void="http://rdfs.org/ns/void#" xmlns:xsd="http://www.w3.org/2001/XMLSchema#" > <rdf:Description rdf:about="https://kops.uni-konstanz.de/server/rdf/resource/123456789/57373"> <dc:creator>Duerr, Oliver</dc:creator> <dc:creator>Hong, Chi Lap</dc:creator> <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/57373"/> <dc:contributor>Scharpf, Philipp</dc:contributor> <foaf:homepage rdf:resource="http://localhost:8080/"/> <dc:language>eng</dc:language> <dcterms:issued>2021</dcterms:issued> <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/> <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2022-04-29T08:19:27Z</dc:date> <dc:creator>Scharpf, Philipp</dc:creator> <dc:contributor>Hong, Chi Lap</dc:contributor> <dcterms:title>Accelerating Active Learning Image Labeling Through Bulk Shift Recommendations</dcterms:title> <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2022-04-29T08:19:27Z</dcterms:available> <dc:contributor>Duerr, Oliver</dc:contributor> <dcterms:abstract xml:lang="eng">Nowadays, the inexpensive memory space promotes an accelerating growth of stored image data. To exploit the data using supervised Machine or Deep Learning, it needs to be labeled. Manually labeling the vast amount of data is time-consuming and expensive, especially if human experts with specific domain knowledge are indispensable. Active learning addresses this shortcoming by querying the user the labels of the most informative images first. One way to obtain the ‘informativeness’ is by using uncertainty sampling as a query strategy, where the system queries those images it is most uncertain about how to classify. In this paper, we present a web-based active learning framework that helps to accelerate the labeling process. After manually labeling some images, the user gets recommendations of further candidates that could potentially be labeled equally (bulk image folder shift). We aim to explore the most efficient ‘uncertainty’ measure to improve the quality of the recommendations such that all images are sorted with a minimum number of user interactions (clicks). We conducted experiments using a manually labeled reference dataset to evaluate different combinations of classifiers and uncertainty measures. The results clearly show the effectiveness of an uncertainty sampling with bulk image shift recommendations (our novel method), which can reduce the number of required clicks to only around 20% compared to manual labeling.</dcterms:abstract> </rdf:Description> </rdf:RDF>