匹配还是不匹配：重新审视图像匹配在可靠视觉位置识别中的作用
To Match or Not to Match: Revisiting Image Matching for Reliable Visual Place Recognition
AI总结：
本研究重新审视视觉位置识别中的图像匹配重排序，发现现代检索系统在数据集饱和时重排序可能有害，提出用内点计数预测重排序收益，以构建更稳健自适应的VPR系统。
摘要：
视觉位置识别（VPR）是计算机视觉中的一项关键任务，传统上通过图像匹配对检索结果进行重排序来增强性能。然而，近期VPR方法的进步显著提升了性能，对重排序的必要性提出了挑战。在本工作中，我们表明现代检索系统常常达到一个临界点，此时重排序反而会降低结果质量，因为当前的VPR数据集在很大程度上已经饱和。我们提出利用图像匹配作为验证步骤来评估检索置信度，证明内点计数能够可靠地预测何时重排序是有益的。我们的发现转变了检索流程的范式，为构建更稳健和自适应的VPR系统提供了见解。代码可在 https://github.com/FarInHeight/To-Match-or-Not-to-Match 获取。
英文摘要：
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. The code is available at https://github.com/FarInHeight/To-Match-or-Not-to-Match.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
