Publikation: Lighting Image/Video Style Transfer Methods by Iterative Channel Pruning
Dateien
Datum
Autor:innen
Herausgeber:innen
ISSN der Zeitschrift
Electronic ISSN
ISBN
Bibliografische Daten
Verlag
Schriftenreihe
Auflagebezeichnung
URI (zitierfähiger Link)
DOI (zitierfähiger Link)
Internationale Patentnummer
Link zur Lizenz
Angaben zur Forschungsförderung
Projekt
Open Access-Veröffentlichung
Core Facility der Universität Konstanz
Titel in einer weiteren Sprache
Publikationstyp
Publikationsstatus
Erschienen in
Zusammenfassung
Deploying style transfer methods on resource-constrained devices is challenging, which limits their real-world applicability. To tackle this issue, we propose using pruning techniques to accelerate various visual style transfer methods. We argue that typical pruning methods may not be well-suited for style transfer methods and present an iterative correlation-based channel pruning (ICCP) strategy for encoder-transform-decoder-based image/video style transfer models. The correlation-based channel regularization preserves the feature distributions for content and style references, and the iterative pruning strategy prevents layer collapse when pruning on the encoder-decoder structure. Experiments demonstrate that the proposed ICCP can generate visual competitive results compared to SOTA style transfer methods and significantly reduces the number of parameters (at least 70K) and inference time. Model is available at https://github.com/wukx-wukx/ICCP.
Zusammenfassung in einer weiteren Sprache
Fachgebiet (DDC)
Schlagwörter
Konferenz
Rezension
Zitieren
ISO 690
WU, Kexin, Fan TANG, Ning LIU, Oliver DEUSSEN, Thi-Ngoc-Hanh LE, Weiming DONG, Tong-Yee LEE, 2024. Lighting Image/Video Style Transfer Methods by Iterative Channel Pruning. ICASSP 2024 : IEEE International Conference on Acoustics, Speech and Signal Processing. Seoul, Korea, 14. Apr. 2024 - 19. Apr. 2024. In: ICASSP 2024 : IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Proceedings. Piscataway, NJ: IEEE, 2024, pp. 3800-3804. ISSN 1520-6149. eISSN 2379-190X. ISBN 979-8-3503-4485-1. Available under: doi: 10.1109/icassp48485.2024.10446950BibTex
@inproceedings{Wu2024Light-69650, year={2024}, doi={10.1109/icassp48485.2024.10446950}, title={Lighting Image/Video Style Transfer Methods by Iterative Channel Pruning}, isbn={979-8-3503-4485-1}, issn={1520-6149}, publisher={IEEE}, address={Piscataway, NJ}, booktitle={ICASSP 2024 : IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Proceedings}, pages={3800--3804}, author={Wu, Kexin and Tang, Fan and Liu, Ning and Deussen, Oliver and Le, Thi-Ngoc-Hanh and Dong, Weiming and Lee, Tong-Yee} }
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/123456789/69650"> <dcterms:hasPart rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/69650/1/Wu_2-1m528lc3rvpg8.PDF"/> <dc:contributor>Tang, Fan</dc:contributor> <dcterms:available rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-03-20T09:55:37Z</dcterms:available> <dspace:isPartOfCollection rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <bibo:uri rdf:resource="https://kops.uni-konstanz.de/handle/123456789/69650"/> <dcterms:isPartOf rdf:resource="https://kops.uni-konstanz.de/server/rdf/resource/123456789/36"/> <dcterms:abstract>Deploying style transfer methods on resource-constrained devices is challenging, which limits their real-world applicability. To tackle this issue, we propose using pruning techniques to accelerate various visual style transfer methods. We argue that typical pruning methods may not be well-suited for style transfer methods and present an iterative correlation-based channel pruning (ICCP) strategy for encoder-transform-decoder-based image/video style transfer models. The correlation-based channel regularization preserves the feature distributions for content and style references, and the iterative pruning strategy prevents layer collapse when pruning on the encoder-decoder structure. Experiments demonstrate that the proposed ICCP can generate visual competitive results compared to SOTA style transfer methods and significantly reduces the number of parameters (at least 70K) and inference time. Model is available at https://github.com/wukx-wukx/ICCP.</dcterms:abstract> <dspace:hasBitstream rdf:resource="https://kops.uni-konstanz.de/bitstream/123456789/69650/1/Wu_2-1m528lc3rvpg8.PDF"/> <dc:creator>Deussen, Oliver</dc:creator> <dc:creator>Wu, Kexin</dc:creator> <dc:contributor>Dong, Weiming</dc:contributor> <dc:creator>Tang, Fan</dc:creator> <dc:contributor>Deussen, Oliver</dc:contributor> <foaf:homepage rdf:resource="http://localhost:8080/"/> <dc:contributor>Le, Thi-Ngoc-Hanh</dc:contributor> <dc:date rdf:datatype="http://www.w3.org/2001/XMLSchema#dateTime">2024-03-20T09:55:37Z</dc:date> <dcterms:title>Lighting Image/Video Style Transfer Methods by Iterative Channel Pruning</dcterms:title> <dc:contributor>Lee, Tong-Yee</dc:contributor> <dcterms:issued>2024</dcterms:issued> <dc:contributor>Liu, Ning</dc:contributor> <dc:creator>Le, Thi-Ngoc-Hanh</dc:creator> <dc:contributor>Wu, Kexin</dc:contributor> <void:sparqlEndpoint rdf:resource="http://localhost/fuseki/dspace/sparql"/> <dc:creator>Liu, Ning</dc:creator> <dc:creator>Lee, Tong-Yee</dc:creator> <dc:creator>Dong, Weiming</dc:creator> <dc:language>eng</dc:language> </rdf:Description> </rdf:RDF>