SOTAVerified

Image Matching

Papers

Showing 110 of 18 papers

TitleStatusHype
GIM: Learning Generalizable Image Matcher From Internet VideosCode4
LightGlue: Local Feature Matching at Light SpeedCode4
RoMa: Robust Dense Feature MatchingCode3
Shared Coupling-bridge for Weakly Supervised Local Feature LearningCode0
Efficient Linear Attention for Fast and Accurate Keypoint Matching0
DKM: Dense Kernelized Feature Matching for Geometry EstimationCode2
Decoupling Makes Weakly Supervised Local Feature BetterCode1
HarrisZ^+: Harris Corner Selection for Next-Gen Image Matching Pipelines0
LoFTR: Detector-Free Local Feature Matching with TransformersCode1
DISK: Learning local features with policy gradientCode1
Show:102550
← PrevPage 1 of 2Next →

Benchmark Results

#ModelMetricClaimedVerifiedStatus
1GIM-RoMaMean AUC@5°53.3Unverified
2GIM-DKMMean AUC@5°51.2Unverified
3RoMaMean AUC@5°48.8Unverified
4DKMMean AUC@5°46.2Unverified
5GIM-LoFTRMean AUC@5°39.1Unverified
6GIM-LightGlueMean AUC@5°38.3Unverified
7LoFTRMean AUC@5°33.1Unverified
8RootSIFTMean AUC@5°31.8Unverified
9LightGlueMean AUC@5°31.7Unverified
10SuperGlueMean AUC@5°31.2Unverified
#ModelMetricClaimedVerifiedStatus
1Harris Cornermean average accuracy @ 100.66Unverified
2DISKmean average accuracy @ 100.65Unverified
3SuperGluemean average accuracy @ 100.65Unverified
4DoG-AffNet-HardNet8mean average accuracy @ 100.64Unverified
5Key.Net-SOSNetmean average accuracy @ 100.6Unverified
6RootSIFTmean average accuracy @ 100.6Unverified
7R2D2mean average accuracy @ 100.56Unverified
8D2-Net (MS)mean average accuracy @ 100.36Unverified