GFNet: Homography Estimation via Grid Flow Regression

Kaining Zhang, Jiayi Ma, Paolo Favaro

Keywords: homography estimation, multimodal, image matching

Abstract: Current deep homography estimation methods are constrained to processing image pairs with limited resolution due to restrictions in network architecture and computational capacity. For larger images, downsampling is often necessary, which can significantly degrade estimation accuracy. To address this limitation, we propose GFNet, a Grid Flow regression Network that consistently delivers high-accuracy homography estimates across varying image resolutions. Unlike previous methods that directly regress the parameters of the global homography between two views, GFNet directly estimates flow over a coarse grid and then uses the resulting correspondences to compute the homography. This approach not only supports high-resolution processing but also preserves the high accuracy of dense matching while significantly reducing the computational load typically associated with such frameworks, thanks to the use of coarse grid flow. We demonstrate the effectiveness of GFNet on a wide range of experiments on multiple datasets, including the common scene MSCOCO, multimodal datasets VIS-IR and GoogleMap, and the dynamic scene VIRAT. In specific, on GoogleMap, GFNet achieves an improvement of +9.9\% in auc@3 while reducing MACs by $\sim$47\% compared to the SOTA dense matching method. Additionally, it shows a 1.7$\times$ improvement in auc@3 over the SOTA deep homography method.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 4703