DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences
Background: Exploring the correspondences across multi-view images is the basis of many computer vision tasks. However, most existing methods are limited on accuracy under challenging conditions. In order to learn more robust and accurate correspondences, we propose the DSD-MatchingNet for local fea...
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
KeAi Communications Co., Ltd.
2022-10-01
|
| Series: | Virtual Reality & Intelligent Hardware |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2096579622000821 |
| _version_ | 1797991333830328320 |
|---|---|
| author | Yicheng Zhao Han Zhang Ping Lu Ping Li EnHua Wu Bin Sheng |
| author_facet | Yicheng Zhao Han Zhang Ping Lu Ping Li EnHua Wu Bin Sheng |
| author_sort | Yicheng Zhao |
| collection | DOAJ |
| description | Background: Exploring the correspondences across multi-view images is the basis of many computer vision tasks. However, most existing methods are limited on accuracy under challenging conditions. In order to learn more robust and accurate correspondences, we propose the DSD-MatchingNet for local feature matching in this paper. First, we develop a deformable feature extraction module to obtain multi-level feature maps, which harvests contextual information from dynamic receptive fields. The dynamic receptive fields provided by deformable convolution network ensures our method to obtain dense and robust correspondences. Second, we utilize the sparse-to-dense matching with the symmetry of correspondence to implement accurate pixel-level matching, which enables our method to produce more accurate correspondences. Experiments have shown that our proposed DSD-MatchingNet achieves a better performance on image matching benchmark, as well as on visual localization benchmark. Specifically, our method achieves 91.3% mean matching accuracy on HPatches dataset and 99.3% visual localization recalls on Aachen Day-Night dataset. |
| first_indexed | 2024-04-11T08:50:28Z |
| format | Article |
| id | doaj.art-3a36445eff1246ed8901a2816078fc74 |
| institution | Directory Open Access Journal |
| issn | 2096-5796 |
| language | English |
| last_indexed | 2024-04-11T08:50:28Z |
| publishDate | 2022-10-01 |
| publisher | KeAi Communications Co., Ltd. |
| record_format | Article |
| series | Virtual Reality & Intelligent Hardware |
| spelling | doaj.art-3a36445eff1246ed8901a2816078fc742022-12-22T04:33:35ZengKeAi Communications Co., Ltd.Virtual Reality & Intelligent Hardware2096-57962022-10-0145432443DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate CorrespondencesYicheng Zhao0Han Zhang1Ping Lu2Ping Li3EnHua Wu4Bin Sheng5Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, ChinaZTE Corporation, China; State Key Laboratory of Mobile Network and Mobile Multimedia Technology, ChinaZTE Corporation, China; State Key Laboratory of Mobile Network and Mobile Multimedia Technology, China; Corresponding author.Department of Computing, The Hong Kong Polytechnic University, Hong Kong; School of Design, The Hong Kong Polytechnic University, Hong KongState Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences, Beijing, China; Faculty of Science and Technology, University of Macau, Macau, ChinaDepartment of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai, China; Corresponding author.Background: Exploring the correspondences across multi-view images is the basis of many computer vision tasks. However, most existing methods are limited on accuracy under challenging conditions. In order to learn more robust and accurate correspondences, we propose the DSD-MatchingNet for local feature matching in this paper. First, we develop a deformable feature extraction module to obtain multi-level feature maps, which harvests contextual information from dynamic receptive fields. The dynamic receptive fields provided by deformable convolution network ensures our method to obtain dense and robust correspondences. Second, we utilize the sparse-to-dense matching with the symmetry of correspondence to implement accurate pixel-level matching, which enables our method to produce more accurate correspondences. Experiments have shown that our proposed DSD-MatchingNet achieves a better performance on image matching benchmark, as well as on visual localization benchmark. Specifically, our method achieves 91.3% mean matching accuracy on HPatches dataset and 99.3% visual localization recalls on Aachen Day-Night dataset.http://www.sciencedirect.com/science/article/pii/S2096579622000821Image matchingDeformable convolution networkSparse-to-dense matching |
| spellingShingle | Yicheng Zhao Han Zhang Ping Lu Ping Li EnHua Wu Bin Sheng DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences Virtual Reality & Intelligent Hardware Image matching Deformable convolution network Sparse-to-dense matching |
| title | DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences |
| title_full | DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences |
| title_fullStr | DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences |
| title_full_unstemmed | DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences |
| title_short | DSD-MatchingNet: Deformable Sparse-to-Dense Feature Matching for Learning Accurate Correspondences |
| title_sort | dsd matchingnet deformable sparse to dense feature matching for learning accurate correspondences |
| topic | Image matching Deformable convolution network Sparse-to-dense matching |
| url | http://www.sciencedirect.com/science/article/pii/S2096579622000821 |
| work_keys_str_mv | AT yichengzhao dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences AT hanzhang dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences AT pinglu dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences AT pingli dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences AT enhuawu dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences AT binsheng dsdmatchingnetdeformablesparsetodensefeaturematchingforlearningaccuratecorrespondences |