SuperPlace: The Renaissance of Classical Feature Aggregation for Visual Place Recognition in the Era of Foundation Models

TL;DR

SuperPlace revives classical feature aggregation techniques for visual place recognition, integrating foundation model concepts, and introduces novel training and aggregation strategies that outperform recent methods on benchmark datasets.

Contribution

The paper introduces a unified training framework, a novel G$^2$M aggregation method, and a fine-tuning strategy for NetVLAD, advancing classical methods in the era of foundation models.

Findings

G$^2$M achieves high performance with fewer feature dimensions.

NVL-FT$^2$ ranks first on the MSLS leaderboard.

SuperPlace outperforms recent state-of-the-art methods.

Abstract

Recent visual place recognition (VPR) approaches have leveraged foundation models (FM) and introduced novel aggregation techniques. However, these methods have failed to fully exploit key concepts of FM, such as the effective utilization of extensive training sets, and they have overlooked the potential of classical aggregation methods, such as GeM and NetVLAD. Building on these insights, we revive classical feature aggregation methods and develop more fundamental VPR models, collectively termed SuperPlace. First, we introduce a supervised label alignment method that enables training across various VPR datasets within a unified framework. Second, we propose G$^2$M, a compact feature aggregation method utilizing two GeMs, where one GeM learns the principal components of feature maps along the channel dimension and calibrates the output of the other. Third, we propose the secondary fine-tuning (FT$^2$) strategy for NetVLAD-Linear (NVL). NetVLAD first learns feature vectors in a high-dimensional space and then compresses them into a lower-dimensional space via a single linear layer. Extensive experiments highlight our contributions and demonstrate the superiority of SuperPlace. Specifically, G$^2$M achieves promising results with only one-tenth of the feature dimensions compared to recent methods. Moreover, NVL-FT$^2$ ranks first on the MSLS leaderboard.