PaperCVPR · 2021
LoFTR: Detector-Free Local Feature Matching with Transformers
1page built on it
4more pages cite it
2narratives
1registry paper cites it
In the Atlas
ModelPrimary source of
LoFTR
Detector-free dense feature matcher: shared CNN backbone produces coarse and fine feature maps, a Linear Transformer with interleaved self- and cross-attention establishes confidence-thresholded mutual nearest-neighbour correspondences, and a fine module refines each match to sub-pixel accuracy.
Cited as a reference by 4 pages, grouped by domain
Features 3
Representation learning 1
Appears in 2 narratives
Lineage
Papers in the Atlas registry only · one dot per paperBuilt upon by · 1
Builds on · 4
2004Distinctive Image Features from Scale-Invariant KeypointsLowe2011ORB: An efficient alternative to SIFT or SURFRublee, Rabaud et al.2018SuperPoint: Self-Supervised Interest Point Detection and DescriptionDeTone, Malisiewicz, Rabinovich2020SuperGlue: Learning Feature Matching with Graph Neural NetworksSarlin, DeTone et al.