S2DNet: Learning Accurate Correspondences for Sparse-to-Dense Feature Matching

S2DNet: Learning Accurate Correspondences for Sparse-to-Dense Feature Matching. Germain, H., Bourmaud, G., & Lepetit, V. arXiv:2004.01673 [cs], 4, 2020.

Paper

S2DNet: Learning Accurate Correspondences for Sparse-to-Dense Feature Matching [link]

Website abstract bibtex

Establishing robust and accurate correspondences is a fundamental backbone to many computer vision algorithms. While recent learning-based feature matching methods have shown promising results in providing robust correspondences under challenging conditions, they are often limited in terms of precision. In this paper, we introduce S2DNet, a novel feature matching pipeline, designed and trained to efficiently establish both robust and accurate correspondences. By leveraging a sparse-to-dense matching paradigm, we cast the correspondence learning problem as a supervised classification task to learn to output highly peaked correspondence maps. We show that S2DNet achieves state-of-the-art results on the HPatches benchmark, as well as on several long-term visual localization datasets.

@article{
 title = {S2DNet: Learning Accurate Correspondences for Sparse-to-Dense Feature Matching},
 type = {article},
 year = {2020},
 keywords = {Computer Science - Computer Vision and Pattern Rec},
 websites = {http://arxiv.org/abs/2004.01673},
 month = {4},
 id = {e63e32c1-63e6-3668-9d6a-f1745573af59},
 created = {2022-03-28T09:45:06.461Z},
 accessed = {2022-01-03},
 file_attached = {true},
 profile_id = {235249c2-3ed4-314a-b309-b1ea0330f5d9},
 group_id = {1ff583c0-be37-34fa-9c04-73c69437d354},
 last_modified = {2022-03-29T07:59:36.087Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {true},
 hidden = {false},
 citation_key = {germainS2DNetLearningAccurate2020},
 source_type = {article},
 short_title = {S2DNet},
 notes = {arXiv: 2004.01673},
 private_publication = {false},
 abstract = {Establishing robust and accurate correspondences is a fundamental backbone to many computer vision algorithms. While recent learning-based feature matching methods have shown promising results in providing robust correspondences under challenging conditions, they are often limited in terms of precision. In this paper, we introduce S2DNet, a novel feature matching pipeline, designed and trained to efficiently establish both robust and accurate correspondences. By leveraging a sparse-to-dense matching paradigm, we cast the correspondence learning problem as a supervised classification task to learn to output highly peaked correspondence maps. We show that S2DNet achieves state-of-the-art results on the HPatches benchmark, as well as on several long-term visual localization datasets.},
 bibtype = {article},
 author = {Germain, Hugo and Bourmaud, Guillaume and Lepetit, Vincent},
 journal = {arXiv:2004.01673 [cs]}
}

Downloads: 0