To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition
Abstract
Visual Place Recognition (VPR) is a critical task in computer vision, traditionally enhanced by re-ranking retrieval results with image matching. However, recent advancements in VPR methods have significantly improved performance, challenging the necessity of re-ranking. In this work, we show that modern retrieval systems often reach a point where re-ranking can degrade results, as current VPR datasets are largely saturated. We propose using image matching as a verification step to assess retrieval confidence, demonstrating that inlier counts can reliably predict when re-ranking is beneficial. Our findings shift the paradigm of retrieval pipelines, offering insights for more robust and adaptive VPR systems.
Cite
Text
Sferrazza et al. "To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2025.Markdown
[Sferrazza et al. "To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition." IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, 2025.](https://mlanthology.org/cvprw/2025/sferrazza2025cvprw-match/)BibTeX
@inproceedings{sferrazza2025cvprw-match,
title = {{To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition}},
author = {Sferrazza, Davide and Berton, Gabriele Moreno and Trivigno, Gabriele and Masone, Carlo},
booktitle = {IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops},
year = {2025},
pages = {2849-2860},
url = {https://mlanthology.org/cvprw/2025/sferrazza2025cvprw-match/}
}