Unsupervised Deep Graph Matching Based on Cycle Consistency

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Tourani, Siddharth, Rother, Carsten, Khan, Muhammad Haris, Savchynskyy, Bogdan
Format: Preprint
Publié: 2023
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914673103732736
author Tourani, Siddharth
Rother, Carsten
Khan, Muhammad Haris
Savchynskyy, Bogdan
author_facet Tourani, Siddharth
Rother, Carsten
Khan, Muhammad Haris
Savchynskyy, Bogdan
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2307_08930
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Unsupervised Deep Graph Matching Based on Cycle Consistency
Tourani, Siddharth
Rother, Carsten
Khan, Muhammad Haris
Savchynskyy, Bogdan
Computer Vision and Pattern Recognition
Artificial Intelligence
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.
title Unsupervised Deep Graph Matching Based on Cycle Consistency
topic Computer Vision and Pattern Recognition
Artificial Intelligence
url https://arxiv.org/abs/2307.08930