KappaPlace: Learning Hyperspherical Uncertainty for Visual Place Recognition via Prototype-Anchored Supervision

2026-05-19

AI Summary

This work addresses the critical limitation of existing visual place recognition methods—their lack of reliable and well-calibrated uncertainty estimates, which undermines decision-making in safety-critical scenarios when handling ambiguous queries or erroneous matches. To this end, we propose KappaPlace, a novel framework that introduces, for the first time, a prototype-anchored supervision strategy. It models image descriptors as von Mises–Fisher distributions and employs a lightweight module to predict concentration parameters that capture epistemic uncertainty. Crucially, KappaPlace enables uncertainty quantification at the match level rather than merely at the query level. The approach supports both joint training and post-hoc extension with frozen backbones, achieving up to a 50% reduction in expected calibration error (ECE@K) across five benchmark datasets while maintaining or improving recall, thereby substantially enhancing uncertainty calibration quality and overall system reliability.

Abstract

Visual Place Recognition (VPR) is critical for autonomous navigation, yet state-of-the-art methods lack well-calibrated uncertainty estimation. Standard pipelines cannot reliably signal when a query is ambiguous or a match is likely incorrect, posing risks in safety-critical robotics. We propose KappaPlace, a principled framework for learning uncertainty-aware VPR representations. Our core contribution is a Prototype-Anchored supervision strategy that leverages latent class representatives as targets for a probabilistic objective. By modeling image descriptors as von Mises-Fisher (vMF) variables, we learn a lightweight module to predict the concentration parameter as a direct proxy for aleatoric uncertainty. While existing VPR uncertainty methods are typically restricted to a query-centric view, we derive a novel match-level formulation to quantify the reliability of specific query-reference pairs. Across five diverse benchmarks, KappaPlace reduces Expected Calibration Error (ECE@K) by up to 50% compared to existing methods while maintaining or improving retrieval recall. We provide both a joint-training variant and a post-training extension for frozen backbones. Our results demonstrate that KappaPlace provides a robust, stable, and well-calibrated signal that enables reliable decision-making within the VPR pipeline. Our code is available at: https://github.com/mayayank95/UncertaintyAwareVPR

Problem

• Visual Place Recognition
• Uncertainty Estimation
• Aleatoric Uncertainty
• Calibration
• Autonomous Navigation

Innovation

• Prototype-Anchored Supervision
• von Mises-Fisher Distribution
• Aleatoric Uncertainty
• Match-Level Uncertainty
• Visual Place Recognition