mlsanityData to 7 Oct 2026

Matching 2D Images in 3D: Metric Relative Pose from Metric Correspondences

arXiv 2404.06337 ↗First posted 9 Apr 2024cs.CVcited by 25

Not on the map. Why?

Abstract (arXiv)

Given two images, we can estimate the relative camera pose between them by establishing image-to-image correspondences. Usually, correspondences are 2D-to-2D and the pose we estimate is defined only up to scale. Some applications, aiming at instant augmented reality anywhere, require scale-metric pose estimates, and hence, they rely on external depth estimators to recover the scale. We present MicKey, a keypoint matching pipeline that is able to predict metric correspondences in 3D camera space. By learning to match 3D coordinates across images, we are able to infer the metric relative pose without depth measurements. Depth measurements are also not required for training, nor are scene reconstructions or image overlap information. MicKey is supervised only by pairs of images and their relative poses. MicKey achieves state-of-the-art performance on the Map-Free Relocalisation benchmark while requiring less supervision than competing approaches.

Key facts. Cited by 25 papers in the map (data through 7 Oct 2026). Builds on 62 of its 94 references that are in the map. First posted on arXiv 9 Apr 2024.

Builds on

62 of 94 references.

References · 62

  1. 2023DINOv2: Learning Robust Visual Features without Supervisioncited by 7,764
  2. 2015Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shiftcited by 6,979
  3. 2017ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenescited by 3,088
  4. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,708
  5. 2020Transformers are RNNs: Fast Autoregressive Transformers with Linear Attentioncited by 1,979
  6. 2021Vision Transformers for Dense Predictioncited by 1,789
  7. 2019SuperGlue: Learning Feature Matching with Graph Neural Networkscited by 1,437
  8. 2017SuperPoint: Self-Supervised Interest Point Detection and Descriptioncited by 1,313
  9. 2014Depth Map Prediction from a Single Image using a Multi-Scale Deep Networkcited by 1,264
  10. 2021LoFTR: Detector-Free Local Feature Matching with Transformerscited by 974
  11. 2016Unsupervised Monocular Depth Estimation with Left-Right Consistencycited by 819
  12. 2018MegaDepth: Learning Single-View Depth Prediction from Internet Photoscited by 731
38 more
  1. 2023ZoeDepth: Zero-shot Transfer by Combining Relative and Metric Depthcited by 646
  2. 2023LightGlue: Local Feature Matching at Light Speedcited by 583
  3. 2020Curriculum Learning for Reinforcement Learning Domains: A Framework and Surveycited by 350
  4. 2018GeoNet: Unsupervised Learning of Dense Depth, Optical Flow and Camera Posecited by 341
  5. 2020DISK: Learning local features with policy gradientcited by 310
  6. 2016DSAC - Differentiable RANSAC for Camera Localizationcited by 264
  7. 2019DIODE: A Dense Indoor and Outdoor DEpth Datasetcited by 236
  8. 2017Geometric Loss Functions for Camera Pose Regression with Deep Learningcited by 224
  9. 2020Image Matching across Wide Baselines: From Paper to Practicecited by 224
  10. 2019R2D2: Repeatable and Reliable Detector and Descriptorcited by 206
  11. 2021COTR: Correspondence Transformer for Matching Across Imagescited by 182
  12. 2017Learning to Find Good Correspondencescited by 176
  13. 2020Visual Camera Re-Localization from RGB and RGB-D Images Using DSACcited by 176
  14. 2022ASpanFormer: Detector-Free Image Matching with Adaptive Span Transformercited by 170
  15. 2017CNN-SLAM: Real-time dense monocular SLAM with learned depth predictioncited by 152
  16. 2018DF-Net: Unsupervised Joint Learning of Depth and Flow using Cross-Task Consistencycited by 148
  17. 2019Learning Two-View Correspondences and Geometry Using Order-Aware Networkcited by 148
  18. 2017Learning Less is More - 6D Camera Localization via 3D Surface Regressioncited by 146
  19. 2019SOSNet: Second Order Similarity Regularization for Local Descriptor Learningcited by 131
  20. 2019Neural-Guided RANSAC: Learning Where to Sample Model Hypothesescited by 129
  21. 2020ASLFeat: Learning Local Features of Accurate Shape and Localizationcited by 129
  22. 2019Self-Supervised Monocular Depth Hintscited by 127
  23. 2019Key.Net: Keypoint Detection by Handcrafted and Learned CNN Filterscited by 117
  24. 2021Learning Accurate Dense Correspondences and When to Trust Themcited by 113
  25. 2022Map-free Visual Relocalization: Metric Pose Relative to a Single Imagecited by 112
  26. 2021Learning Multi-Scene Absolute Pose Regression with Transformerscited by 108
  27. 2018PlaneRCNN: 3D Plane Detection and Reconstruction from a Single Imagecited by 105
  28. 2019Hierarchical Scene Coordinate Classification and Regression for Visual Localizationcited by 93
  29. 2023DeDoDe: Detect, Don't Describe -- Describe, Don't Detect for Local Feature Matchingcited by 62
  30. 2014TILDE: A Temporally Invariant Learned DEtectorcited by 55
  31. 2023SiLK -- Simple Learned Keypointscited by 54
  32. 2021Wide-Baseline Relative Camera Pose Estimation with Directional Learningcited by 49
  33. 2021Extreme Rotation Estimation using Dense Correlation Volumescited by 42
  34. 2019End-to-End CAD Model Retrieval and 9DoF Alignment in 3D Scanscited by 40
  35. 2019Reinforced Feature Points: Optimizing Feature Detection and Description for a High-Level Taskcited by 39
  36. 2022Generalized Differentiable RANSACcited by 30
  37. 2019NM-Net: Mining Reliable Neighbors for Robust Feature Correspondencescited by 29
  38. 2022Decoupling Makes Weakly Supervised Local Feature Bettercited by 28

50 most cited of 62 All 62 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 5; 2025: 14; 2026: 6.

25 papers.

Sort:

  1. 2024PF3plat: Pose-Free Feed-Forward 3D Gaussian Splattingcited by 60
  2. 2025DUNE: Distilling a Universal Encoder from Heterogeneous 2D and 3D Teacherscited by 28
  3. 2025FG2: Fine-Grained Cross-View Localization by Fine-Grained Feature Matchingcited by 27
  4. 2025GaSLight: Gaussian Splats for Spatially-Varying Lighting in HDRcited by 11
  5. 2025Rig3R: Rig-Aware Conditioning for Learned 3D Reconstructioncited by 11
  6. 2024Multiview Scene Graphcited by 9
  7. 2025Loc2: Interpretable Cross-View Localization via Depth-Lifted Local Feature Matchingcited by 8
  8. 2025ACE-G: Improving Generalization of Scene Coordinate Regression Through Query Pre-Trainingcited by 8
  9. 2024Symmetry Strikes Back: From Single-Image Symmetry Detection to 3D Generationcited by 7
  10. 2024Can Generative Video Models Help Pose Estimation?cited by 7
  11. 2025CoMatcher: Multi-View Collaborative Feature Matchingcited by 7
  12. 2024Mismatched: Evaluating the Limits of Image Matching Approaches and Benchmarkscited by 4
13 more
  1. 2025A Scene is Worth a Thousand Features: Feed-Forward Camera Localization from a Collection of Image Featurescited by 4
  2. 2025SegMASt3R: Geometry Grounded Segment Matchingcited by 4
  3. 2025What does really matter in image goal navigation?cited by 2
  4. 2025Modality-Aware Feature Matching in Visual and Vision-Language Applications: A Comprehensive Surveycited by 2
  5. 2025ViewBridge:Revisiting Cross-View Localization from Image Matchingcited by 2
  6. 2025Scene Coordinate Reconstruction Priorscited by 2
  7. 2025Kinaema: a recurrent sequence model for memory and pose in motioncited by 2
  8. 2026VVLoc: Prior-free 3-DoF Vehicle Visual Localizationcited by 0
  9. 2026Reference-Induced Consensus for Selective Posed-Reference Visual Localizationcited by 0
  10. 2026PIXIE: A Zero-Shot texture-invariant 6D pose estimation framework for unseen objects with assembly defectscited by 0
  11. 2026RIPE++: Reinforced Keypoint Learning from Positive Pairs Onlycited by 0
  12. 2026AutoCompass: Accurate Visual Localization on Public Maps by Learning from Weak Labelscited by 0
  13. 2026ARC-Loc: Leveraging Azimuthal Ray Convergence as a Geometric Cue for Direct Cross-View Localizationcited by 0