English

Unsupervised Deep Graph Matching Based on Cycle Consistency

Computer Vision and Pattern Recognition 2024-02-13 v5 Artificial Intelligence

Abstract

We contribute to the sparsely populated area of unsupervised deep graph matching with application to keypoint matching in images. Contrary to the standard \emph{supervised} approach, our method does not require ground truth correspondences between keypoint pairs. Instead, it is self-supervised by enforcing consistency of matchings between images of the same object category. As the matching and the consistency loss are discrete, their derivatives cannot be straightforwardly used for learning. We address this issue in a principled way by building our method upon the recent results on black-box differentiation of combinatorial solvers. This makes our method exceptionally flexible, as it is compatible with arbitrary network architectures and combinatorial solvers. Our experimental evaluation suggests that our technique sets a new state-of-the-art for unsupervised graph matching.

Keywords

Cite

@article{arxiv.2307.08930,
  title  = {Unsupervised Deep Graph Matching Based on Cycle Consistency},
  author = {Siddharth Tourani and Carsten Rother and Muhammad Haris Khan and Bogdan Savchynskyy},
  journal= {arXiv preprint arXiv:2307.08930},
  year   = {2024}
}

Comments

12 pages, 5 figures, 3 papers

R2 v1 2026-06-28T11:33:07.424Z