Towards Test-time Efficient Visual Place Recognition via Asymmetric Query Processing

2025-12-15

AI Summary

Visual place recognition (VPR) on resource-constrained devices faces prohibitive online inference overhead when deploying high-capacity foundation models (e.g., DINOv2). This paper proposes a test-time efficient asymmetric VPR framework: gallery features are extracted offline using DINOv2, while a lightweight query network enables real-time matching during inference. Our key contributions are: (1) a novel geographically structured memory bank that replaces conventional k-NN retrieval and supports end-to-end differentiable training; and (2) an implicit embedding enhancement mechanism that improves the robustness of small query networks to viewpoint and illumination variations. Evaluated on multiple standard benchmarks, our method achieves state-of-the-art accuracy–efficiency trade-offs at significantly lower computational cost, outperforming existing asymmetric VPR approaches.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Hardware-aware ML
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Visual Place Recognition (VPR) has advanced significantly with high-capacity foundation models like DINOv2, achieving remarkable performance. Nonetheless, their substantial computational cost makes deployment on resource-constrained devices impractical. In this paper, we introduce an efficient asymmetric VPR framework that incorporates a high-capacity gallery model for offline feature extraction with a lightweight query network for online processing. A key challenge in this setting is ensuring compatibility between these heterogeneous networks, which conventional approaches address through computationally expensive k-NN-based compatible training. To overcome this, we propose a geographical memory bank that structures gallery features using geolocation metadata inherent in VPR databases, eliminating the need for exhaustive k-NN computations. Additionally, we introduce an implicit embedding augmentation technique that enhances the query network to model feature variations despite its limited capacity. Extensive experiments demonstrate that our method not only significantly reduces computational costs but also outperforms existing asymmetric retrieval techniques, establishing a new aspect for VPR in resource-limited environments. The code is available at https://github.com/jaeyoon1603/AsymVPR

Problem

• Reduces computational cost for visual place recognition on devices
• Ensures compatibility between high-capacity and lightweight networks efficiently
• Enhances query network capacity to model feature variations effectively

Innovation

• Asymmetric framework with lightweight query network
• Geographical memory bank eliminating k-NN computations
• Implicit embedding augmentation for limited capacity modeling