cs.CVFeb 25, 2026

Automatic Map Density Selection for Locally-Performant Visual Place Recognition

Authors: Somayeh Hussaini, Tobias Fischer, Michael Milford

Organizations: QUT Centre for Robotics, Queensland University of Technology, Brisbane, QLD 4000, Australia

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

A key challenge in translating Visual Place Recognition (VPR) from the lab to long-term deployment is ensuring a priori that a system can meet user-specified performance requirements across different parts of an environment, rather than just on average globally. One critical mechanism for controlling this local performance is the density of the reference mapping database, yet this factor is largely neglected in existing work, where fixed, engineering-driven sampling densities based on sensors, storage, or GPS frequency are typically used. In this paper, we propose a VPR mapping approach that uses two reference traverses from the operating environment to automatically select an appropriate map density satisfying two user-defined requirements: (1) a target Local Recall@1 level, and (2) the proportion of the operational environment over which it must be met or exceeded, which we term the Recall Achievement Rate (RAR). Our approach uses spatial consistency and coherence features from reference-to-reference matches at multiple map densities to estimate the density needed to meet these targets on unseen deployment data. Through experiments across two VPR methods and the Nordland and Oxford RobotCar benchmarks, we show that our system consistently meets or exceeds the target Local Recall@1 over at least the user-specified proportion of the environment. Comparisons with alternative baselines show that it reliably selects an appropriate operating point in map density, avoiding unnecessarily dense maps. Finally, ablations evaluate sensitivity to reference traversal choice and segment length, and our analysis reveals that conventional global Recall@1 is a poor predictor of the often more operationally meaningful RAR metric.

Figures & tables

Fig. 1 : Our approach models the relationship between segment-wise Recall@1 and features extracted from matching two reference traverses at multiple candidate map densities. Given a user-specified Local Recall@1 threshold and target Recall Achievement Rate (RAR), it selects either a uniform or per-segment reference density predicted to satisfy the threshold over the required proportion of environmental segments.
Fig. 2 : Overview of our automatic reference density selection framework. (A) Two aligned reference traversals are segmented and processed at multiple candidate densities. (B) Features extracted from the Ref1-to-Ref2 distance matrix are used to train density-specific Ridge regressors that predict segment-level Recall@1. (C) Given user-specified Local Recall@1 and Recall Achievement Rate (RAR) targets, the selected policy determines either a single constant factor k∗k^{\ast} or per-segment factors {ki∗}\{k_{i}^{\ast}\} . (D) The resulting reference map Ref1 ∗ is evaluated against the unseen query traversal Qry1.
Fig. 3 : Success rate versus average reference ratio across all target Local Recall@1 and Recall Achievement Rate (RAR) combinations, VPR methods, and datasets. Success rate is the proportion of configurations for which the achieved RAR meets or exceeds its target. The reference ratio is the retained reference count normalised by the densest map at k=1k=1 . Circles denote fixed global densities, while the star, triangle, and pentagon denote the constant, variable, and budget-equalised variable selectors, respectively.
TABLE I : Achieved RAR values. Top block : achieved RAR per cell; blue = above target RAR threshold, red = below. ∗* indicates the cell is at its maximum achievable RAR (ceiling) despite falling below the target threshold. Bottom block : reference count ratio normalised by kk =1; green = sparsest (fewest refs), pale yellow = equal to densest ( kk =1). Last column: mean ref ratio (normalised by kk =1) across RAR thresholds.
Fig. 4 : Qualitative examples of the selected reference density under two user-specified operating points. Selected reference frames are shown above the query sequence; green and red borders indicate correct and incorrect top-1 matches, respectively. [Top] For a target Local Recall@1 of 60% and target RAR of 40%, the method selects k=15k=15 , and the illustrated segment achieves 60% Recall@1. [Bottom] For a target Local Recall@1 of 80% and target RAR of 60%, the method selects k=5k=5 , and the illustrated segment achieves 90% Recall@1.
Fig. 5 : Recall Achievement Rate (RAR) as a function of the target Local Recall@1 threshold for different fixed reference densities, from dense ( k=1k=1 ) to sparse ( k=50k=50 ), using CosPlace on Nordland. For each threshold, RAR is the proportion of segments whose Local Recall@1 meets or exceeds that threshold.
Fig. 6 : Relationship between mean Recall@1 and Recall Achievement Rate (RAR) for fixed reference densities k∈Kk\in\mathcal{K} under Local Recall@1 targets of 20%, 40%, 60%, 80%, and 100%, using MixVPR on Nordland. For a fixed density, mean Recall@1 remains unchanged, while RAR decreases as the Local Recall@1 target becomes stricter because fewer segments satisfy the requirement.
Fig. 7 : Effect of Ridge-regression regularisation strength λ\lambda on selector success rate across both VPR methods and datasets. Performance remains stable for λ∈[10−3,1]\lambda\in[10^{-3},1] and degrades under stronger regularisation. The default value λ=0.01\lambda=0.01 , used in all experiments, lies within this stable region.
More in this topic

Visual Place Recognition

Visual Place Recognition matches camera observations to mapped places despite changes in viewpoint, lighting, weather, or time.

See all 84 papers

Explore similar work

CardsList
Jul 14, 2026

Breaking Déjà Vu: Independent Auditing of Visual Place Recognition through Vision-Language Reasoning

Visual place recognition (VPR) is a key enabler of accurate localization and long-term autonomous navigation in robotics applications, such as loop closure detection for simultaneous localisation and mapping (SLAM). However, real-world VPR deployment relies on selecting an image matching threshold that balances precision and recall. These thresholds are typically tuned using labeled validation data and fixed during deployment, making them unreliable under environmental changes where ground truth is unavailable. This is particularly problematic in safety-critical robotics, where accepting a false loop closure can corrupt the estimated trajectory and map. In this work, we introduce Visual Place Recognition Auditing, an independent post-retrieval verification framework that leverages Vision-Language Models (VLMs) to assess retrieved matches by reasoning jointly over query and candidate images. Unlike conventional verification methods, our approach performs instance-level verification without requiring architecture-specific confidence measures, dataset-dependent thresholds, or prior knowledge of the deployment environment. We evaluate our method on six benchmark datasets using five state-of-the-art VPR methods and four VLMs. Results show that VLM-based auditing improves recall@1 by 13.6% on average as compared to state-of-the-art methods while reducing false acceptance rates to 12%, maintaining precision above 95% and coverage above 75%.

Sania Waheed, Michael Milford, Sarvapali D. Ramchurn +1

School of Electronics and Computer Science, University of Southampton, Southampton SO17 1BJ, U.K. · School of Electrical Engineering and Computer Science, Queensland University of Technology, Brisbane, QLD 4000, Australia · School of Computer Science and Electronic Engineering, University of Essex, CO4 3SQ Colchester, U.K.

Visual Place RecognitionRobot Localization
AskPDF
Mar 14, 2026

Evaluation of Visual Place Recognition Methods for Image Pair Retrieval in 3D Vision and Robotics

Visual Place Recognition (VPR) is a core component in computer vision, typically formulated as an image retrieval task for localization, mapping, and navigation. In this work, we instead study VPR as an image pair retrieval front-end for registration pipelines, where the goal is to find top-matching image pairs between two disjoint image sets for downstream tasks such as scene registration, SLAM, and Structure-from-Motion. We comparatively evaluate state-of-the-art VPR families - NetVLAD-style baselines, classification-based global descriptors (CosPlace, EigenPlaces), feature-mixing (MixVPR), and foundation-model-driven methods (AnyLoc, SALAD, MegaLoc) - on three challenging datasets: object-centric outdoor scenes (Tanks and Temples), indoor RGB-D scans (ScanNet-GS), and autonomous-driving sequences (KITTI). We show that modern global descriptor approaches are increasingly suitable as off-the-shelf image pair retrieval modules in challenging scenarios including perceptual aliasing and incomplete sequences, while exhibiting clear, domain-dependent strengths and weaknesses that are critical when choosing VPR components for robust mapping and registration.

Dennis Haitz, Athradi Shritish Shetty, Michael Weinmann +1

Karlsruhe Institute of Technology, Institute of Photogrammetry and Remote Sensing, Karlsruhe, Germany · Delft University of Technology, Department of Intelligent Systems, The Netherlands

Visual Place RecognitionImage Retrieval
AskPDF
Aug 7, 2026

Are Visual Place Recognition Models Recognizing Places or Conditions? Distractor-Augmented Evaluation and Condition Suppression

Long-term Visual Place Recognition (VPR) is typically evaluated by matching queries from one condition against a database from another. Crowdsourced map databases, however, may mix conditions and include images that resemble the query in condition but depict different places. In the presence of these distractors, a method may retrieve by condition similarity rather than place identity. We argue that this susceptibility arises because the discriminability of VPR methods allows them to encode information such as illumination, weather, and seasonal appearance in their descriptors. We therefore introduce Distractor-Augmented Recall (DAR) to isolate and quantify the effect of distractors, and propose condition suppression to remove condition information from VPR descriptors. Across eleven methods and six datasets, method rankings under DAR@1 differ from those under Recall@1 (R@1), while applying INLP and LEACE as condition suppression methods generally improves DAR@1 without reducing R@1. Thus, distractor robustness is distinct from standard retrieval performance and can be improved by suppressing condition information.

Beomsu Kim, Minwoo Jung, Giseop Kim

Department of Robotics and Mechatronics Engineering, DGIST, Daegu, Republic of Korea · Department of Mechanical Engineering, Seoul National University, Seoul, Republic of Korea

Visual Place RecognitionRetrieval RobustnessInvariant Representation LearningConcept Erasure
AskPDF