Lost in the Tail: Addressing Geographic Imbalance in Urban Visual Place Recognition

2026-06-30

AI Summary

This study addresses the long-tailed geographical distribution of visual place recognition (VPR) data at city scale, which biases models toward frequently sampled regions while neglecting sparsely covered areas. The work presents the first systematic characterization of this geographical imbalance and introduces DAPR—a plug-and-play, model-agnostic framework that enhances recognition performance in underrepresented regions. DAPR achieves this by balancing optimization between head and tail classes through distribution-aware gradient reweighting, complemented by multi-scale distance search and a unified classification-retrieval mechanism. Evaluated on SF-XL, DAPR improves over existing classification-retrieval baselines by 18.3% (v1) and 6.7% (v2), and consistently boosts performance across multiple benchmarks—including MSLS and Pitts30k—for a range of state-of-the-art VPR methods.

Abstract

Urban-scale Visual Place Recognition (VPR) aims to identify the geographic location of a query image by matching it against a geo-tagged database. While recent methods achieve impressive performance, they overlook a serious long-tailed problem hidden in urban-scale datasets, which biases the model towards locations with abundant images and ignores less-visited areas, causing models to systematically favor frequently photographed locations while failing in sparsely covered areas. In this paper, we systematically characterize this imbalance challenge and propose Distribution-Aware Place Recognition (DAPR), a model-agnostic plug-in framework that rebalances gradient contributions across head and tail classes. Additionally, within classification-retrieval pipelines, DAPR applies a multi-scale distance search mechanism to compute per-class distributional compactness, providing complementary gains at the retrieval stage. On the large-scale SF-XL benchmark, our framework outperforms the previous classification-retrieval baseline by 18.3% on test set v1, and 6.7% on test set v2. As a plug-in module, it achieves consistent improvements across representative VPR methods on SF-XL, MSLS, and Pitts30k, demonstrating broad generalizability across different methods and benchmarks.

Problem

• Geographic Imbalance
• Long-tailed Distribution
• Visual Place Recognition
• Urban-scale VPR
• Data Imbalance

Innovation

• long-tailed distribution
• geographic imbalance
• visual place recognition
• distribution-aware learning
• gradient rebalancing