File Download

There are no files associated with this item.

  Links for fulltext
     (May Require Subscription)
Supplementary

Conference Paper: Multi-scale Matching Networks for Semantic Correspondence

TitleMulti-scale Matching Networks for Semantic Correspondence
Authors
KeywordsComputer vision
Codes
Fuses
Semantics
Buildings
Issue Date2021
PublisherIEEE Computer Society.
Citation
ICCV Workshop on Deep Multi-Task Learning in Computer Vision (Virtual), Montreal, QC, Canada, October 11-17, 2021. In Proceedings: 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW 2021), p. 3334-3344 How to Cite?
AbstractDeep features have been proven powerful in building accurate dense semantic correspondences in various previous works. However, the multi-scale and pyramidal hierarchy of convolutional neural networks has not been well studied to learn discriminative pixel-level features for semantic correspondence. In this paper, we propose a multi-scale matching network that is sensitive to tiny semantic differences between neighboring pixels. We follow the coarse-to-fine matching strategy and build a top-down feature and matching enhancement scheme that is coupled with the multi-scale hierarchy of deep convolutional neural networks. During feature enhancement, intra-scale enhancement fuses same-resolution feature maps from multiple layers together via local self-attention and cross-scale enhancement hallucinates higher-resolution feature maps along the top-down pathway. Besides, we learn complementary matching details at different scales thus the overall matching score is refined by features of different semantic levels gradually. Our multi-scale matching network can be trained end-to-end easily with few additional learnable parameters. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on three popular benchmarks with high computational efficiency. The code has been released at https://github.com/wintersun661/MMNet.
Persistent Identifierhttp://hdl.handle.net/10722/316359
ISI Accession Number ID

 

DC FieldValueLanguage
dc.contributor.authorZhao, D-
dc.contributor.authorSong, Z-
dc.contributor.authorJi, Z-
dc.contributor.authorZhao, G-
dc.contributor.authorGe, W-
dc.contributor.authorYu, Y-
dc.date.accessioned2022-09-02T06:10:06Z-
dc.date.available2022-09-02T06:10:06Z-
dc.date.issued2021-
dc.identifier.citationICCV Workshop on Deep Multi-Task Learning in Computer Vision (Virtual), Montreal, QC, Canada, October 11-17, 2021. In Proceedings: 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW 2021), p. 3334-3344-
dc.identifier.urihttp://hdl.handle.net/10722/316359-
dc.description.abstractDeep features have been proven powerful in building accurate dense semantic correspondences in various previous works. However, the multi-scale and pyramidal hierarchy of convolutional neural networks has not been well studied to learn discriminative pixel-level features for semantic correspondence. In this paper, we propose a multi-scale matching network that is sensitive to tiny semantic differences between neighboring pixels. We follow the coarse-to-fine matching strategy and build a top-down feature and matching enhancement scheme that is coupled with the multi-scale hierarchy of deep convolutional neural networks. During feature enhancement, intra-scale enhancement fuses same-resolution feature maps from multiple layers together via local self-attention and cross-scale enhancement hallucinates higher-resolution feature maps along the top-down pathway. Besides, we learn complementary matching details at different scales thus the overall matching score is refined by features of different semantic levels gradually. Our multi-scale matching network can be trained end-to-end easily with few additional learnable parameters. Experimental results demonstrate that the proposed method achieves state-of-the-art performance on three popular benchmarks with high computational efficiency. The code has been released at https://github.com/wintersun661/MMNet.-
dc.languageeng-
dc.publisherIEEE Computer Society.-
dc.relation.ispartofProceedings: 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW 2021)-
dc.rightsProceedings: 2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW 2021). Copyright © IEEE Computer Society.-
dc.subjectComputer vision-
dc.subjectCodes-
dc.subjectFuses-
dc.subjectSemantics-
dc.subjectBuildings-
dc.titleMulti-scale Matching Networks for Semantic Correspondence-
dc.typeConference_Paper-
dc.identifier.emailYu, Y: yzyu@cs.hku.hk-
dc.identifier.authorityYu, Y=rp01415-
dc.identifier.doi10.1109/ICCV48922.2021.00334-
dc.identifier.hkuros336342-
dc.identifier.spage3334-
dc.identifier.epage3344-
dc.identifier.isiWOS:000797698903054-
dc.publisher.placeUnited States-

Export via OAI-PMH Interface in XML Formats


OR


Export to Other Non-XML Formats