Full search page →

Learned features in more detail

Optional · supplementary note 10 of 10 for IP 12 · Feature Extraction. Nothing in the main text depends on it; citation numbers refer to the reference list at the bottom of this page.

Navigate: this page and this track: Navigate: this page and this track

Learned methods now cover every stage of local-feature matching.

Learned keypoints and descriptors. SuperPoint [14] trains one fully convolutional network to output both a keypoint heatmap and dense descriptors. It is self-supervised: it first learns corners on synthetic shapes, then improves itself on real images through homographic adaptation, aggregating its own detections over many random warps of the same image. DISK [16] trains detection and description end to end with reinforcement learning (policy gradient), rewarding keypoints that lead to correct matches. These replace the formulas of Harris and SIFT with learned functions, but keep the same idea: sparse, repeatable points with descriptors.

Learned matching. SuperGlue [15] replaces the nearest-neighbor ratio test with a graph neural network that lets keypoints in both images attend to each other, then solves an optimal-transport assignment that can also say “no match”. LightGlue [18] reworks this design to be faster and adaptive, spending less computation on easy image pairs. LoFTR [17] removes the detector altogether: a Transformer with self- and cross-attention matches coarse feature maps densely and then refines positions, which helps in low-texture areas where detectors find nothing.

General-purpose deep features. The deepest change is that features are increasingly not designed for a task at all. Self-supervised foundation models such as DINOv2 [19] produce patch features that work across classification, segmentation, depth estimation and retrieval without fine-tuning. RoMa [20] shows the effect on matching: it uses frozen DINOv2 features for coarse matching, combined with fine convolutional features for precise localization. In other words, the handcrafted descriptor at the center of SIFT has been replaced by a network trained on a very large image collection.

References

  1. R. C. Gonzalez and R. E. Woods, Digital Image Processing, 4th ed., Pearson, 2018, Ch. 12. publisher page
  2. M.-K. Hu, “Visual pattern recognition by moment invariants,” IRE Transactions on Information Theory, vol. 8, no. 2, pp. 179–187, 1962. doi
  3. R. M. Haralick, K. Shanmugam, and I. Dinstein, “Textural features for image classification,” IEEE Transactions on Systems, Man, and Cybernetics, vol. SMC-3, no. 6, pp. 610–621, 1973. doi
  4. C. Harris and M. Stephens, “A combined corner and edge detector,” in Proc. Alvey Vision Conference, pp. 23.1–23.6, 1988. doi
  5. J. Matas, O. Chum, M. Urban, and T. Pajdla, “Robust wide-baseline stereo from maximally stable extremal regions,” Image and Vision Computing, vol. 22, no. 10, pp. 761–767, 2004. doi
  6. D. G. Lowe, “Distinctive image features from scale-invariant keypoints,” International Journal of Computer Vision, vol. 60, no. 2, pp. 91–110, 2004. doi
  7. H. Bay, A. Ess, T. Tuytelaars, and L. Van Gool, “Speeded-up robust features (SURF),” Computer Vision and Image Understanding, vol. 110, no. 3, pp. 346–359, 2008. doi
  8. E. Rublee, V. Rabaud, K. Konolige, and G. Bradski, “ORB: An efficient alternative to SIFT or SURF,” in Proc. IEEE International Conference on Computer Vision (ICCV), pp. 2564–2571, 2011. doi
  9. D. Zhang and G. Lu, “Review of shape representation and description techniques,” Pattern Recognition, vol. 37, no. 1, pp. 1–19, 2004. doi
  10. T. Tuytelaars and K. Mikolajczyk, “Local invariant feature detectors: A survey,” Foundations and Trends in Computer Graphics and Vision, vol. 3, no. 3, pp. 177–280, 2008. doi
  11. L. Liu, J. Chen, P. Fieguth, G. Zhao, R. Chellappa, and M. Pietikäinen, “From BoW to CNN: Two decades of texture representation for texture classification,” International Journal of Computer Vision, vol. 127, pp. 74–109, 2019. arXiv
  12. J. Ma, X. Jiang, A. Fan, J. Jiang, and J. Yan, “Image matching from handcrafted to deep features: A survey,” International Journal of Computer Vision, vol. 129, pp. 23–79, 2021. doi
  13. Y. Jin, D. Mishkin, A. Mishchuk, J. Matas, P. Fua, K. M. Yi, and E. Trulls, “Image matching across wide baselines: From paper to practice,” International Journal of Computer Vision, 2021. doi · arXiv
  14. D. DeTone, T. Malisiewicz, and A. Rabinovich, “SuperPoint: Self-supervised interest point detection and description,” arXiv:1712.07629, 2017. arXiv
  15. P.-E. Sarlin, D. DeTone, T. Malisiewicz, and A. Rabinovich, “SuperGlue: Learning feature matching with graph neural networks,” arXiv:1911.11763, 2019. arXiv
  16. M. J. Tyszkiewicz, P. Fua, and E. Trulls, “DISK: Learning local features with policy gradient,” arXiv:2006.13566, 2020. arXiv
  17. J. Sun, Z. Shen, Y. Wang, H. Bao, and X. Zhou, “LoFTR: Detector-free local feature matching with Transformers,” arXiv:2104.00680, 2021. arXiv
  18. P. Lindenberger, P.-E. Sarlin, and M. Pollefeys, “LightGlue: Local feature matching at light speed,” arXiv:2306.13643, 2023. arXiv
  19. M. Oquab, T. Darcet, T. Moutakanni, et al., “DINOv2: Learning robust visual features without supervision,” arXiv:2304.07193, 2023. arXiv
  20. J. Edstedt, Q. Sun, G. Bökman, M. Wadenbäck, and M. Felsberg, “RoMa: Robust dense feature matching,” in Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2024. arXiv
  21. M. A. Fischler and R. C. Bolles, “Random sample consensus: A paradigm for model fitting with applications to image analysis and automated cartography,” Communications of the ACM, vol. 24, no. 6, pp. 381–395, 1981. doi
  22. scikit-image contributors, “skimage.measure.ransac,” scikit-image 0.26 documentation (API reference). docs