Dual-level Progressive Hardness-Aware Reweighting for Cross-View Geo-Localization

2025-10-31

AI Summary

Cross-view geolocalization (CVGL) faces severe viewpoint discrepancies and hard negative sample interference between drone and satellite imagery; existing static reweighting strategies often induce gradient noise and training instability. To address this, we propose a dual-level progressive hardness-aware reweighting framework: (i) a ratio-based hardness-aware module that dynamically evaluates relative sample difficulty at the instance level; and (ii) a training-progress-aware progressive adaptive loss weighting mechanism that smoothly modulates hard negative weights at the batch level. This design jointly ensures early-stage training stability and late-stage discriminative capability, effectively suppressing gradient noise while enhancing hard sample mining. Our method achieves significant improvements over state-of-the-art approaches on the University-1652 and SUES-200 benchmarks, demonstrating both superior localization accuracy and robustness to viewpoint and semantic mismatches.

Technology Category

• Computer Vision: Learning & Optimization for CV
• Machine Learning: Multi-instance/Multi-view Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Data and user privacy-enhancing technologies for the Web

Abstract

Cross-view geo-localization (CVGL) between drone and satellite imagery remains challenging due to severe viewpoint gaps and the presence of hard negatives, which are visually similar but geographically mismatched samples. Existing mining or reweighting strategies often use static weighting, which is sensitive to distribution shifts and prone to overemphasizing difficult samples too early, leading to noisy gradients and unstable convergence. In this paper, we present a Dual-level Progressive Hardness-aware Reweighting (DPHR) strategy. At the sample level, a Ratio-based Difficulty-Aware (RDA) module evaluates relative difficulty and assigns fine-grained weights to negatives. At the batch level, a Progressive Adaptive Loss Weighting (PALW) mechanism exploits a training-progress signal to attenuate noisy gradients during early optimization and progressively enhance hard-negative mining as training matures. Experiments on the University-1652 and SUES-200 benchmarks demonstrate the effectiveness and robustness of the proposed DPHR, achieving consistent improvements over state-of-the-art methods.

Problem

• Addresses viewpoint gaps in drone-satellite cross-view geo-localization
• Mitigates hard negative interference in cross-view matching tasks
• Improves training stability and convergence through progressive reweighting

Innovation

• Dual-level progressive hardness-aware reweighting strategy
• Ratio-based difficulty-aware module assigns fine-grained weights
• Progressive adaptive loss weighting attenuates noisy gradients